Re: [MonetDB-users] 64bit MonetDB, JDBC Insert via RJDBC, >300 million rows
Now have tried the same bulk load with COPY on the latest (v5.10.0) with single thread setting. Failed again! System: OS X MonetDB complied in 64bit number of files to load: 322 total size of files: 15GB Max row in file: 3million On Wed, Mar 18, 2009 at 11:51 AM, Martin Kersten <Martin.Kersten@cwi.nl>wrote:
Yue Sheng wrote:
I'm not sure how "The (parallel) load used scratch area as well" is related to the question.
If you look at the code, you will notice that there is a two phase loading process involving (possibly) multiple threads
Sorry if I'm a bit slow.
On Wed, Mar 18, 2009 at 11:25 AM, Martin Kersten <Martin.Kersten@cwi.nl<mailto: Martin.Kersten@cwi.nl>> wrote:
Yue Sheng wrote:
Sorry, if I wasn't clear on the first question:
(1) we ramp up N for the first insert to claim sufficient space. Sure, understand that one.
But:
The claimed space got "given back" *right after* the first insert. (this is the part I don't understand.)
The (parallel) load used scratch area as well
Question: how does the second, third, .... inserts get the "benefit" of the ramp up that we did for the first insert?
Is this a bit clearer what my question pertains?
Thanks.
On Wed, Mar 18, 2009 at 10:26 AM, Martin Kersten <Martin.Kersten@cwi.nl <mailto:Martin.Kersten@cwi.nl> <mailto:Martin.Kersten@cwi.nl <mailto:Martin.Kersten@cwi.nl>>> wrote:
Yue Sheng wrote:
Three questions that bothers me are: (1) why we need to ramp up N to total of all line in first insert.
to let the kernel claim sufficient space
Reason I ask is that right after first insert, the allocation drop right down from, say 100GB to 35GB, and stays roughly there for *all* subsequent inserts. I totally do not understand this. (2) in your opinion, based on this experience, what could be the potential problem here?
little to none, as the files are memory mapped, which only may cause io on some systems
(3) in your opinion, would the newer version cure the problem?
a system can never correctly guess what will come, especially since the source of a COPY command need not be a file but standard input, i.e. a stream.
Thanks.
On Tue, Mar 17, 2009 at 10:51 PM, Martin Kersten <Martin.Kersten@cwi.nl <mailto:Martin.Kersten@cwi.nl> <mailto:Martin.Kersten@cwi.nl <mailto:Martin.Kersten@cwi.nl>> <mailto:Martin.Kersten@cwi.nl <mailto:Martin.Kersten@cwi.nl> <mailto:Martin.Kersten@cwi.nl <mailto:Martin.Kersten@cwi.nl>>>>
wrote:
Yue Sheng wrote:
Martin,
It almost worked...
This is what I did and what have happened:
I have 322 files to insert into data base, totaling 650 million rows
I divided the file list into two, then for each sub list
(a) I insert first file in the list with N set to 650milllion rows, (b) all subsequent files have N set to the number of lines in *that* file
once first list done, then
(c) I insert first file in the second list with N set to 650million rows, (d) all subsequent files have N set to the number of lines in *that* file
Then the same problem happened: it stucked at file number 316.
ok. using the 650M enables MonetDB to allocate enough space and does not have to fall back on guessing. Guessing is painful, because when a file of N records has been created and it needs more, it makes a file of size 1.3xN. This leads to memory fragmentation.
in your case i would have been a little mode spacious and used 700M as a start, because miscalculation of 1 gives a lot of pain. Such advice is only needed in (a)
Note: This is farther then previous tries, which all stopped in the region of file number 280 +/- a few.
My observation: (i) at (a), the VSIZE went up to around 46GB, then after first insert, it drops to around 36GB
ok fits
(ii) at (c), the VSIZE went up to around 130GB, then after first insert, it drops to around 45GB
you tell the system to extend existing BATs prepare for another 650 M, which means it allocates 2*36 G, plus room for the old one gives 108GB then during processings some temporary BATs may be needed,e.g. to check integrity constraints after each file,. Then it runs out of swapspace.
(iii) the "Free Memory", as reported by Activity Monitor, just before it failed at file number 316, dipped to as low as 7.5MB!
yes, you are running out of swapspace on your system. This should not have happened, because the system uses mmapped files and may be an issue with the MacOS or relate to a problem we fixed recently
My question: (1) why we need to ramp N up to total number of lines (it takes along time to do that), then only have it drop down to 30GB-40GB right after
this might indicate that on MacOS, just like Windows, mmaped files need to be written to disk. With a disk bandwidth of 50MB/sec it still takes several minutes
the first insert and stay roughly there? Does it mean we're giving back all the pre-allocation space back to the OS? Then should we set N always to total number of lines? If so, it would take much much longer to process all the files... (2) How come RSIZE never goes above 4GB? (3) Does sql log file size have some limit, that we need to tweak?
no limit
(4) Has anyone successfully implemented the 64bit version of MondeDB and successfully inserted more than 1billion rows?
you platform may be the first, but Amherst has worked with Macs for years
(5) when you say you "...The VSIZE of 44G is not too problematic, i am looking at queries letting it tumble between 20-80 GB....," What does it mean? Mine went up to as high as 135GB...
explained above.
regards, Martin
Thanks, as always.
Managed to COPY 161 files to one table and another 162 files to another table. But when I do a simple select count(*) from table2; it failed (nothing, just hangs).... Has anyone successfully run this database on the MAC, in 64bit, with table (not database) size larger than 500million rows (10 columns)? On Thu, Mar 19, 2009 at 8:59 AM, Yue Sheng <yuesheng8@gmail.com> wrote:
Now have tried the same bulk load with COPY on the latest (v5.10.0) with single thread setting. Failed again!
System: OS X MonetDB complied in 64bit number of files to load: 322 total size of files: 15GB Max row in file: 3million
On Wed, Mar 18, 2009 at 11:51 AM, Martin Kersten <Martin.Kersten@cwi.nl>wrote:
Yue Sheng wrote:
I'm not sure how "The (parallel) load used scratch area as well" is related to the question.
If you look at the code, you will notice that there is a two phase loading process involving (possibly) multiple threads
Sorry if I'm a bit slow.
On Wed, Mar 18, 2009 at 11:25 AM, Martin Kersten <Martin.Kersten@cwi.nl<mailto: Martin.Kersten@cwi.nl>> wrote:
Yue Sheng wrote:
Sorry, if I wasn't clear on the first question:
(1) we ramp up N for the first insert to claim sufficient space. Sure, understand that one.
But:
The claimed space got "given back" *right after* the first insert. (this is the part I don't understand.)
The (parallel) load used scratch area as well
Question: how does the second, third, .... inserts get the "benefit" of the ramp up that we did for the first insert?
Is this a bit clearer what my question pertains?
Thanks.
On Wed, Mar 18, 2009 at 10:26 AM, Martin Kersten <Martin.Kersten@cwi.nl <mailto:Martin.Kersten@cwi.nl> <mailto:Martin.Kersten@cwi.nl <mailto:Martin.Kersten@cwi.nl>>> wrote:
Yue Sheng wrote:
Three questions that bothers me are: (1) why we need to ramp up N to total of all line in first insert.
to let the kernel claim sufficient space
Reason I ask is that right after first insert, the allocation drop right down from, say 100GB to 35GB, and stays roughly there for *all* subsequent inserts. I totally do not understand this. (2) in your opinion, based on this experience, what could be the potential problem here?
little to none, as the files are memory mapped, which only may cause io on some systems
(3) in your opinion, would the newer version cure the problem?
a system can never correctly guess what will come, especially since the source of a COPY command need not be a file but standard input, i.e. a stream.
Thanks.
On Tue, Mar 17, 2009 at 10:51 PM, Martin Kersten <Martin.Kersten@cwi.nl <mailto:Martin.Kersten@cwi.nl> <mailto:Martin.Kersten@cwi.nl <mailto:Martin.Kersten@cwi.nl>> <mailto:Martin.Kersten@cwi.nl <mailto:Martin.Kersten@cwi.nl> <mailto:Martin.Kersten@cwi.nl <mailto:Martin.Kersten@cwi.nl>>>>
wrote:
Yue Sheng wrote:
Martin,
It almost worked...
This is what I did and what have happened:
I have 322 files to insert into data base, totaling 650 million rows
I divided the file list into two, then for each sub list
(a) I insert first file in the list with N set to 650milllion rows, (b) all subsequent files have N set to the number of lines in *that* file
once first list done, then
(c) I insert first file in the second list with N set to 650million rows, (d) all subsequent files have N set to the number of lines in *that* file
Then the same problem happened: it stucked at file number 316.
ok. using the 650M enables MonetDB to allocate enough space and does not have to fall back on guessing. Guessing is painful, because when a file of N records has been created and it needs more, it makes a file of size 1.3xN. This leads to memory fragmentation.
in your case i would have been a little mode spacious and used 700M as a start, because miscalculation of 1 gives a lot of pain. Such advice is only needed in (a)
Note: This is farther then previous tries, which all stopped in the region of file number 280 +/- a few.
My observation: (i) at (a), the VSIZE went up to around 46GB, then after first insert, it drops to around 36GB
ok fits
(ii) at (c), the VSIZE went up to around 130GB, then after first insert, it drops to around 45GB
you tell the system to extend existing BATs prepare for another 650 M, which means it allocates 2*36 G, plus room for the old one gives 108GB then during processings some temporary BATs may be needed,e.g. to check integrity constraints after each file,. Then it runs out of swapspace.
(iii) the "Free Memory", as reported by Activity Monitor, just before it failed at file number 316, dipped to as low as 7.5MB!
yes, you are running out of swapspace on your system. This should not have happened, because the system uses mmapped files and may be an issue with the MacOS or relate to a problem we fixed recently
My question: (1) why we need to ramp N up to total number of lines (it takes along time to do that), then only have it drop down to 30GB-40GB right after
this might indicate that on MacOS, just like Windows, mmaped files need to be written to disk. With a disk bandwidth of 50MB/sec it still takes several minutes
the first insert and stay roughly there? Does it mean we're giving back all the pre-allocation space back to the OS? Then should we set N always to total number of lines? If so, it would take much much longer to process all the files... (2) How come RSIZE never goes above 4GB? (3) Does sql log file size have some limit, that we need to tweak?
no limit
(4) Has anyone successfully implemented the 64bit version of MondeDB and successfully inserted more than 1billion rows?
you platform may be the first, but Amherst has worked with Macs for years
(5) when you say you "...The VSIZE of 44G is not too problematic, i am looking at queries letting it tumble between 20-80 GB....," What does it mean? Mine went up to as high as 135GB...
explained above.
regards, Martin
Thanks, as always.
We did a gdb after the database "hangs" during the COPY... INTO.... phase (note the server was set to run in SINGLE thread mode, i.e. set gdk_nr_threads = 1): All of the threads are suspect, in the sense that they point out either to a deadlock situation or 'mmap' related issues which are both quite difficult to debug. gdb's "info threads" gives: 6 process 41456 thread 0x2603 0x00007fff846c6d7e in semaphore_wait_signal_trap () 5 process 41456 thread 0x1403 0x00007fff846c6d7e in semaphore_wait_signal_trap () 4 process 41456 thread 0x1203 0x00007fff84745886 in msync () 3 process 41456 thread 0x1103 0x00007fff84712526 in select$DARWIN_EXTSN () 2 process 41456 thread 0xf03 0x00007fff846cdd02 in __semwait_signal () * 1 process 41456 thread 0x10b 0x00007fff846cdd02 in __semwait_signal () The backtrace for each thread is: (gdb) thread apply all bt Thread 6 (process 41456 thread 0x2603): #0 0x00007fff846c6d7e in semaphore_wait_signal_trap () #1 0x00007fff846ce698 in pthread_mutex_lock () #2 0x0000000109dc00fe in store_lock () #3 0x0000000109d8375c in mvc_create () #4 0x0000000109d3e8fa in SQLinitClient (c=0x100115db0) at sql_scenario.c:315 #5 0x000000010005c2c9 in runPhase (c=0x100115db0, phase=5) at mal_scenario.c:352 #6 0x000000010005c332 in runScenarioBody (c=0x100115db0) at mal_scenario.c:375 #7 0x000000010005c5ea in runScenario (c=0x100115db0) at mal_scenario.c:420 #8 0x0000000100017d35 in MSserveClient (dummy=0x100115db0) at mal_session.c:360 #9 0x00007fff846f4dcb in _pthread_start () #10 0x00007fff846f4c8d in thread_start () Thread 5 (process 41456 thread 0x1403): #0 0x00007fff846c6d7e in semaphore_wait_signal_trap () #1 0x00007fff846ce698 in pthread_mutex_lock () #2 0x0000000109dc00fe in store_lock () #3 0x0000000109d82ad1 in mvc_trans () #4 0x0000000109d3ebe1 in SQLcacheRefresh (m=0x101872208) at sql_scenario.c:386 #5 0x0000000109d3fb00 in SQLparser (c=0x100115b18) at sql_scenario.c:788 #6 0x000000010005c2c9 in runPhase (c=0x100115b18, phase=1) at mal_scenario.c:352 #7 0x000000010005c3c0 in runScenarioBody (c=0x100115b18) at mal_scenario.c:386 #8 0x000000010005c5ea in runScenario (c=0x100115b18) at mal_scenario.c:420 #9 0x0000000100017d35 in MSserveClient (dummy=0x100115b18) at mal_session.c:360 #10 0x00007fff846f4dcb in _pthread_start () #11 0x00007fff846f4c8d in thread_start () Thread 4 (process 41456 thread 0x1203): #0 0x00007fff84745886 in msync () #1 0x0000000100c38190 in MT_msync (p=0x9ac507000, off=0, len=4391849608, mode=16384) at gdk_posix.c:800 #2 0x0000000100b0ea7c in GDKsave (nme=0x101331d78 "11/1150", ext=0x100d64104 "tail", buf=0x9ac507000, size=4391849608, mode=1) at gdk_storage.c:211 #3 0x000000010089fcc6 in HEAPsave (h=0x10a8ffdb0, nme=0x101331d78 "11/1150", ext=0x100d64104 "tail") at gdk_heap.c:296 #4 0x0000000100b10dfd in BATsave (bd=0x10a02f808) at gdk_storage.c:470 #5 0x000000010089d348 in BBPsync (cnt=250, subcommit=0x10a8fff20) at gdk_bbp.c:2905 #6 0x0000000100886ce7 in TMsubcommit (b=0xc0e057458) at gdk_tm.c:115 #7 0x0000000100c3bc28 in bm_subcommit (list=0x10130ee78, catalog=0x10130ee78, extra=0xc0e07bf38, debug=0) at gdk_logger.c:838 #8 0x0000000100c3eded in bm_commit (lg=0x109f74f08) at gdk_logger.c:1559 #9 0x0000000100c3afc3 in logger_commit (lg=0x109f74f08) at gdk_logger.c:776 #10 0x0000000100c3ca0e in logger_exit (lg=0x109f74f08) at gdk_logger.c:1053 #11 0x0000000100c3cb1d in logger_restart (lg=0x109f74f08) at gdk_logger.c:1080 #12 0x0000000109dca6fd in bl_restart () #13 0x0000000109dc0013 in store_manager () #14 0x0000000109d82a37 in mvc_logmanager () #15 0x00007fff846f4dcb in _pthread_start () #16 0x00007fff846f4c8d in thread_start () Thread 3 (process 41456 thread 0x1103): #0 0x00007fff84712526 in select$DARWIN_EXTSN () #1 0x00000001087f1e35 in SERVERlistenThread (Sock=0x1094ba0f8) at mal_mapi.c:152 #2 0x00007fff846f4dcb in _pthread_start () #3 0x00007fff846f4c8d in thread_start () Thread 2 (process 41456 thread 0xf03): #0 0x00007fff846cdd02 in __semwait_signal () #1 0x00007fff84735f27 in nanosleep () #2 0x0000000100c3882f in MT_sleep_ms (ms=5000) at gdk_posix.c:1774 #3 0x000000010096c45b in GDKvmtrim (limit=0x100d74868) at gdk_utils.c:1350 #4 0x00007fff846f4dcb in _pthread_start () #5 0x00007fff846f4c8d in thread_start () Thread 1 (process 41456 thread 0x10b): #0 0x00007fff846cdd02 in __semwait_signal () #1 0x00007fff84735f27 in nanosleep () #2 0x0000000100c3882f in MT_sleep_ms (ms=5000) at gdk_posix.c:1774 #3 0x0000000100002dd7 in main (argc=7, av=0x7fff5fbfee80) at mserver5.c:514 On Thu, Mar 19, 2009 at 6:01 PM, Yue Sheng <yuesheng8@gmail.com> wrote:
Managed to COPY 161 files to one table and another 162 files to another table. But when I do a simple select count(*) from table2; it failed (nothing, just hangs).... Has anyone successfully run this database on the MAC, in 64bit, with table (not database) size larger than 500million rows (10 columns)?
On Thu, Mar 19, 2009 at 8:59 AM, Yue Sheng <yuesheng8@gmail.com> wrote:
Now have tried the same bulk load with COPY on the latest (v5.10.0) with single thread setting. Failed again!
System: OS X MonetDB complied in 64bit number of files to load: 322 total size of files: 15GB Max row in file: 3million
On Wed, Mar 18, 2009 at 11:51 AM, Martin Kersten <Martin.Kersten@cwi.nl>wrote:
Yue Sheng wrote:
I'm not sure how "The (parallel) load used scratch area as well" is related to the question.
If you look at the code, you will notice that there is a two phase loading process involving (possibly) multiple threads
Sorry if I'm a bit slow.
On Wed, Mar 18, 2009 at 11:25 AM, Martin Kersten <Martin.Kersten@cwi.nl<mailto: Martin.Kersten@cwi.nl>> wrote:
Yue Sheng wrote:
Sorry, if I wasn't clear on the first question:
(1) we ramp up N for the first insert to claim sufficient space. Sure, understand that one.
But:
The claimed space got "given back" *right after* the first insert. (this is the part I don't understand.)
The (parallel) load used scratch area as well
Question: how does the second, third, .... inserts get the "benefit" of the ramp up that we did for the first insert?
Is this a bit clearer what my question pertains?
Thanks.
On Wed, Mar 18, 2009 at 10:26 AM, Martin Kersten <Martin.Kersten@cwi.nl <mailto:Martin.Kersten@cwi.nl> <mailto:Martin.Kersten@cwi.nl <mailto:Martin.Kersten@cwi.nl>>> wrote:
Yue Sheng wrote:
Three questions that bothers me are: (1) why we need to ramp up N to total of all line in first insert.
to let the kernel claim sufficient space
Reason I ask is that right after first insert, the allocation drop right down from, say 100GB to 35GB, and stays roughly there for *all* subsequent inserts. I totally do not understand this. (2) in your opinion, based on this experience, what could be the potential problem here?
little to none, as the files are memory mapped, which only may cause io on some systems
(3) in your opinion, would the newer version cure the problem?
a system can never correctly guess what will come, especially since the source of a COPY command need not be a file but standard input, i.e. a stream.
Thanks.
On Tue, Mar 17, 2009 at 10:51 PM, Martin Kersten <Martin.Kersten@cwi.nl <mailto:Martin.Kersten@cwi.nl> <mailto:Martin.Kersten@cwi.nl <mailto:Martin.Kersten@cwi.nl>> <mailto:Martin.Kersten@cwi.nl <mailto:Martin.Kersten@cwi.nl> <mailto:Martin.Kersten@cwi.nl <mailto:Martin.Kersten@cwi.nl>>>>
wrote:
Yue Sheng wrote:
Martin,
It almost worked...
This is what I did and what have happened:
I have 322 files to insert into data base, totaling 650 million rows
I divided the file list into two, then for each sub list
(a) I insert first file in the list with N set to 650milllion rows, (b) all subsequent files have N set to the number of lines in *that* file
once first list done, then
(c) I insert first file in the second list with N set to 650million rows, (d) all subsequent files have N set to the number of lines in *that* file
Then the same problem happened: it stucked at file number 316.
ok. using the 650M enables MonetDB to allocate enough space and does not have to fall back on guessing. Guessing is painful, because when a file of N records has been created and it needs more, it makes a file of size 1.3xN. This leads to memory fragmentation.
in your case i would have been a little mode spacious and used 700M as a start, because miscalculation of 1 gives a lot of pain. Such advice is only needed in (a)
Note: This is farther then previous tries, which all stopped in the region of file number 280 +/- a few.
My observation: (i) at (a), the VSIZE went up to around 46GB, then after first insert, it drops to around 36GB
ok fits
(ii) at (c), the VSIZE went up to around 130GB, then after first insert, it drops to around 45GB
you tell the system to extend existing BATs prepare for another 650 M, which means it allocates 2*36 G, plus room for the old one gives 108GB then during processings some temporary BATs may be needed,e.g. to check integrity constraints after each file,. Then it runs out of swapspace.
(iii) the "Free Memory", as reported by Activity Monitor, just before it failed at file number 316, dipped to as low as 7.5MB!
yes, you are running out of swapspace on your system. This should not have happened, because the system uses mmapped files and may be an issue with the MacOS or relate to a problem we fixed recently
My question: (1) why we need to ramp N up to total number of lines (it takes along time to do that), then only have it drop down to 30GB-40GB right after
this might indicate that on MacOS, just like Windows, mmaped files need to be written to disk. With a disk bandwidth of 50MB/sec it still takes several minutes
the first insert and stay roughly there? Does it mean we're giving back all the pre-allocation space back to the OS? Then should we set N always to total number of lines? If so, it would take much much longer to process all the files... (2) How come RSIZE never goes above 4GB? (3) Does sql log file size have some limit, that we need to tweak?
no limit
(4) Has anyone successfully implemented the 64bit version of MondeDB and successfully inserted more than 1billion rows?
you platform may be the first, but Amherst has worked with Macs for years
(5) when you say you "...The VSIZE of 44G is not too problematic, i am looking at queries letting it tumble between 20-80 GB....," What does it mean? Mine went up to as high as 135GB...
explained above.
regards, Martin
Thanks, as always.
On Wed, Mar 25, 2009 at 05:02:42PM -0700, Yue Sheng wrote:
We did a gdb after the database "hangs" during the COPY... INTO.... phase (note the server was set to run in SINGLE thread mode, i.e. set gdk_nr_threads = 1):
All of the threads are suspect, in the sense that they point out either to a deadlock situation or 'mmap' related issues which are both quite difficult to debug. gdb's "info threads" gives: 6 process 41456 thread 0x2603 0x00007fff846c6d7e in semaphore_wait_signal_trap ()
5 process 41456 thread 0x1403 0x00007fff846c6d7e in semaphore_wait_signal_trap ()
4 process 41456 thread 0x1203 0x00007fff84745886 in msync ()
3 process 41456 thread 0x1103 0x00007fff84712526 in select$DARWIN_EXTSN ()
2 process 41456 thread 0xf03 0x00007fff846cdd02 in __semwait_signal ()
* 1 process 41456 thread 0x10b 0x00007fff846cdd02 in __semwait_signal () The backtrace for each thread is: (gdb) thread apply all bt Thread 6 (process 41456 thread 0x2603): #0 0x00007fff846c6d7e in semaphore_wait_signal_trap () #1 0x00007fff846ce698 in pthread_mutex_lock () #2 0x0000000109dc00fe in store_lock () #3 0x0000000109d8375c in mvc_create () #4 0x0000000109d3e8fa in SQLinitClient (c=0x100115db0) at sql_scenario.c:315 #5 0x000000010005c2c9 in runPhase (c=0x100115db0, phase=5) at mal_scenario.c:352 #6 0x000000010005c332 in runScenarioBody (c=0x100115db0) at mal_scenario.c:375 #7 0x000000010005c5ea in runScenario (c=0x100115db0) at mal_scenario.c:420 #8 0x0000000100017d35 in MSserveClient (dummy=0x100115db0) at mal_session.c:360 #9 0x00007fff846f4dcb in _pthread_start () #10 0x00007fff846f4c8d in thread_start () Thread 5 (process 41456 thread 0x1403): #0 0x00007fff846c6d7e in semaphore_wait_signal_trap () #1 0x00007fff846ce698 in pthread_mutex_lock () #2 0x0000000109dc00fe in store_lock () #3 0x0000000109d82ad1 in mvc_trans () #4 0x0000000109d3ebe1 in SQLcacheRefresh (m=0x101872208) at sql_scenario.c:386 #5 0x0000000109d3fb00 in SQLparser (c=0x100115b18) at sql_scenario.c:788 #6 0x000000010005c2c9 in runPhase (c=0x100115b18, phase=1) at mal_scenario.c:352 #7 0x000000010005c3c0 in runScenarioBody (c=0x100115b18) at mal_scenario.c:386 #8 0x000000010005c5ea in runScenario (c=0x100115b18) at mal_scenario.c:420 #9 0x0000000100017d35 in MSserveClient (dummy=0x100115b18) at mal_session.c:360 #10 0x00007fff846f4dcb in _pthread_start () #11 0x00007fff846f4c8d in thread_start () The thread 5 and 6 seem to both wait on the same lock. It looks like both threads are seperate clients is this possible, ie are you running multiple mclients or so?
niels
Thread 4 (process 41456 thread 0x1203): #0 0x00007fff84745886 in msync () #1 0x0000000100c38190 in MT_msync (p=0x9ac507000, off=0, len=4391849608, mode=16384) at gdk_posix.c:800 #2 0x0000000100b0ea7c in GDKsave (nme=0x101331d78 "11/1150", ext=0x100d64104 "tail", buf=0x9ac507000, size=4391849608, mode=1) at gdk_storage.c:211 #3 0x000000010089fcc6 in HEAPsave (h=0x10a8ffdb0, nme=0x101331d78 "11/1150", ext=0x100d64104 "tail") at gdk_heap.c:296 #4 0x0000000100b10dfd in BATsave (bd=0x10a02f808) at gdk_storage.c:470 #5 0x000000010089d348 in BBPsync (cnt=250, subcommit=0x10a8fff20) at gdk_bbp.c:2905 #6 0x0000000100886ce7 in TMsubcommit (b=0xc0e057458) at gdk_tm.c:115 #7 0x0000000100c3bc28 in bm_subcommit (list=0x10130ee78, catalog=0x10130ee78, extra=0xc0e07bf38, debug=0) at gdk_logger.c:838 #8 0x0000000100c3eded in bm_commit (lg=0x109f74f08) at gdk_logger.c:1559 #9 0x0000000100c3afc3 in logger_commit (lg=0x109f74f08) at gdk_logger.c:776 #10 0x0000000100c3ca0e in logger_exit (lg=0x109f74f08) at gdk_logger.c:1053 #11 0x0000000100c3cb1d in logger_restart (lg=0x109f74f08) at gdk_logger.c:1080 #12 0x0000000109dca6fd in bl_restart () #13 0x0000000109dc0013 in store_manager () #14 0x0000000109d82a37 in mvc_logmanager () #15 0x00007fff846f4dcb in _pthread_start () #16 0x00007fff846f4c8d in thread_start () Thread 3 (process 41456 thread 0x1103): #0 0x00007fff84712526 in select$DARWIN_EXTSN () #1 0x00000001087f1e35 in SERVERlistenThread (Sock=0x1094ba0f8) at mal_mapi.c:152 #2 0x00007fff846f4dcb in _pthread_start () #3 0x00007fff846f4c8d in thread_start () Thread 2 (process 41456 thread 0xf03): #0 0x00007fff846cdd02 in __semwait_signal () #1 0x00007fff84735f27 in nanosleep () #2 0x0000000100c3882f in MT_sleep_ms (ms=5000) at gdk_posix.c:1774 #3 0x000000010096c45b in GDKvmtrim (limit=0x100d74868) at gdk_utils.c:1350 #4 0x00007fff846f4dcb in _pthread_start () #5 0x00007fff846f4c8d in thread_start () Thread 1 (process 41456 thread 0x10b): #0 0x00007fff846cdd02 in __semwait_signal () #1 0x00007fff84735f27 in nanosleep () #2 0x0000000100c3882f in MT_sleep_ms (ms=5000) at gdk_posix.c:1774 #3 0x0000000100002dd7 in main (argc=7, av=0x7fff5fbfee80) at mserver5.c:514
On Thu, Mar 19, 2009 at 6:01 PM, Yue Sheng <[1]yuesheng8@gmail.com> wrote:
Managed to COPY 161 files to one table and another 162 files to another table. But when I do a simple select count(*) from table2; it failed (nothing, just hangs)....
Has anyone successfully run this database on the MAC, in 64bit, with table (not database) size larger than 500million rows (10 columns)?
On Thu, Mar 19, 2009 at 8:59 AM, Yue Sheng <[2]yuesheng8@gmail.com> wrote:
Now have tried the same bulk load with COPY on the latest (v5.10.0) with single thread setting.
Failed again!
System:
OS X
MonetDB complied in 64bit
number of files to load: 322
total size of files: 15GB
Max row in file: 3million
On Wed, Mar 18, 2009 at 11:51 AM, Martin Kersten <[3]Martin.Kersten@cwi.nl> wrote:
Yue Sheng wrote:
I'm not sure how "The (parallel) load used scratch area as well" is related to the question.
If you look at the code, you will notice that there is a two phase loading process involving (possibly) multiple threads
Sorry if I'm a bit slow.
On Wed, Mar 18, 2009 at 11:25 AM, Martin Kersten <[4]Martin.Kersten@cwi.nl <mailto:[5]Martin.Kersten@cwi.nl>> wrote: Yue Sheng wrote: Sorry, if I wasn't clear on the first question: (1) we ramp up N for the first insert to claim sufficient space. Sure, understand that one. But: The claimed space got "given back" *right after* the first insert. (this is the part I don't understand.) The (parallel) load used scratch area as well Question: how does the second, third, .... inserts get the "benefit" of the ramp up that we did for the first insert? Is this a bit clearer what my question pertains? Thanks. On Wed, Mar 18, 2009 at 10:26 AM, Martin Kersten <[6]Martin.Kersten@cwi.nl <mailto:[7]Martin.Kersten@cwi.nl>
<mailto:[8]Martin.Kersten@cwi.nl <mailto:[9]Martin.Kersten@cwi.nl>>> wrote: Yue Sheng wrote: Three questions that bothers me are: (1) why we need to ramp up N to total of all line in first insert. to let the kernel claim sufficient space Reason I ask is that right after first insert, the allocation drop right down from, say 100GB to 35GB, and stays roughly there for *all* subsequent inserts. I totally do not understand this. (2) in your opinion, based on this experience, what could be the potential problem here? little to none, as the files are memory mapped, which only may cause io on some systems (3) in your opinion, would the newer version cure the problem? a system can never correctly guess what will come, especially since the source of a COPY command need not be a file but standard input, i.e. a stream. Thanks. On Tue, Mar 17, 2009 at 10:51 PM, Martin Kersten <[10]Martin.Kersten@cwi.nl <mailto:[11]Martin.Kersten@cwi.nl> <mailto:[12]Martin.Kersten@cwi.nl <mailto:[13]Martin.Kersten@cwi.nl>> <mailto:[14]Martin.Kersten@cwi.nl <mailto:[15]Martin.Kersten@cwi.nl> <mailto:[16]Martin.Kersten@cwi.nl <mailto:[17]Martin.Kersten@cwi.nl>>>> wrote: Yue Sheng wrote: Martin, It almost worked... This is what I did and what have happened: I have 322 files to insert into data base, totaling 650 million rows I divided the file list into two, then for each sub list (a) I insert first file in the list with N set to 650milllion rows, (b) all subsequent files have N set to the number of lines in *that* file once first list done, then (c) I insert first file in the second list with N set to 650million rows, (d) all subsequent files have N set to the number of lines in *that* file Then the same problem happened: it stucked at file number 316. ok. using the 650M enables MonetDB to allocate enough space and does not have to fall back on guessing. Guessing is painful, because when a file of N records has been created and it needs more, it makes a file of size 1.3xN. This leads to memory fragmentation. in your case i would have been a little mode spacious and used 700M as a start, because miscalculation of 1 gives a lot of pain. Such advice is only needed in (a) Note: This is farther then previous tries, which all stopped in the region of file number 280 +/- a few. My observation: (i) at (a), the VSIZE went up to around 46GB, then after first insert, it drops to around 36GB ok fits (ii) at (c), the VSIZE went up to around 130GB, then after first insert, it drops to around 45GB you tell the system to extend existing BATs prepare for another 650 M, which means it allocates 2*36 G, plus room for the old one gives 108GB then during processings some temporary BATs may be needed,e.g. to check integrity constraints after each file,. Then it runs out of swapspace. (iii) the "Free Memory", as reported by Activity Monitor, just before it failed at file number 316, dipped to as low as 7.5MB! yes, you are running out of swapspace on your system. This should not have happened, because the system uses mmapped files and may be an issue with the MacOS or relate to a problem we fixed recently My question: (1) why we need to ramp N up to total number of lines (it takes along time to do that), then only have it drop down to 30GB-40GB right after this might indicate that on MacOS, just like Windows, mmaped files need to be written to disk. With a disk bandwidth of 50MB/sec it still takes several minutes the first insert and stay roughly there? Does it mean we're giving back all the pre-allocation space back to the OS? Then should we set N always to total number of lines? If so, it would take much much longer to process all the files... (2) How come RSIZE never goes above 4GB? (3) Does sql log file size have some limit, that we need to tweak? no limit (4) Has anyone successfully implemented the 64bit version of MondeDB and successfully inserted more than 1billion rows? you platform may be the first, but Amherst has worked with Macs for years (5) when you say you "...The VSIZE of 44G is not too problematic, i am looking at queries letting it tumble between 20-80 GB....," What does it mean? Mine went up to as high as 135GB... explained above. regards, Martin Thanks, as always.
References
1. mailto:yuesheng8@gmail.com 2. mailto:yuesheng8@gmail.com 3. mailto:Martin.Kersten@cwi.nl 4. mailto:Martin.Kersten@cwi.nl 5. mailto:Martin.Kersten@cwi.nl 6. mailto:Martin.Kersten@cwi.nl 7. mailto:Martin.Kersten@cwi.nl 8. mailto:Martin.Kersten@cwi.nl 9. mailto:Martin.Kersten@cwi.nl 10. mailto:Martin.Kersten@cwi.nl 11. mailto:Martin.Kersten@cwi.nl 12. mailto:Martin.Kersten@cwi.nl 13. mailto:Martin.Kersten@cwi.nl 14. mailto:Martin.Kersten@cwi.nl 15. mailto:Martin.Kersten@cwi.nl 16. mailto:Martin.Kersten@cwi.nl 17. mailto:Martin.Kersten@cwi.nl
------------------------------------------------------------------------------
_______________________________________________ MonetDB-users mailing list MonetDB-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/monetdb-users
-- Niels Nes, Centre for Mathematics and Computer Science (CWI) Science Park 123, 1098 XG Amsterdam, The Netherlands room C0.02, phone ++31 20 592-4098, fax ++31 20 592-4312 url: http://www.cwi.nl/~niels e-mail: Niels.Nes@cwi.nl
On Thu, Mar 26, 2009 at 07:49:49AM +0100, Niels Nes wrote:
On Wed, Mar 25, 2009 at 05:02:42PM -0700, Yue Sheng wrote:
We did a gdb after the database "hangs" during the COPY... INTO.... phase (note the server was set to run in SINGLE thread mode, i.e. set gdk_nr_threads = 1):
All of the threads are suspect, in the sense that they point out either to a deadlock situation or 'mmap' related issues which are both quite difficult to debug. gdb's "info threads" gives: 6 process 41456 thread 0x2603 0x00007fff846c6d7e in semaphore_wait_signal_trap ()
5 process 41456 thread 0x1403 0x00007fff846c6d7e in semaphore_wait_signal_trap ()
4 process 41456 thread 0x1203 0x00007fff84745886 in msync ()
3 process 41456 thread 0x1103 0x00007fff84712526 in select$DARWIN_EXTSN ()
2 process 41456 thread 0xf03 0x00007fff846cdd02 in __semwait_signal ()
* 1 process 41456 thread 0x10b 0x00007fff846cdd02 in __semwait_signal () The backtrace for each thread is: (gdb) thread apply all bt Thread 6 (process 41456 thread 0x2603): #0 0x00007fff846c6d7e in semaphore_wait_signal_trap () #1 0x00007fff846ce698 in pthread_mutex_lock () #2 0x0000000109dc00fe in store_lock () #3 0x0000000109d8375c in mvc_create () #4 0x0000000109d3e8fa in SQLinitClient (c=0x100115db0) at sql_scenario.c:315 #5 0x000000010005c2c9 in runPhase (c=0x100115db0, phase=5) at mal_scenario.c:352 #6 0x000000010005c332 in runScenarioBody (c=0x100115db0) at mal_scenario.c:375 #7 0x000000010005c5ea in runScenario (c=0x100115db0) at mal_scenario.c:420 #8 0x0000000100017d35 in MSserveClient (dummy=0x100115db0) at mal_session.c:360 #9 0x00007fff846f4dcb in _pthread_start () #10 0x00007fff846f4c8d in thread_start () Thread 5 (process 41456 thread 0x1403): #0 0x00007fff846c6d7e in semaphore_wait_signal_trap () #1 0x00007fff846ce698 in pthread_mutex_lock () #2 0x0000000109dc00fe in store_lock () #3 0x0000000109d82ad1 in mvc_trans () #4 0x0000000109d3ebe1 in SQLcacheRefresh (m=0x101872208) at sql_scenario.c:386 #5 0x0000000109d3fb00 in SQLparser (c=0x100115b18) at sql_scenario.c:788 #6 0x000000010005c2c9 in runPhase (c=0x100115b18, phase=1) at mal_scenario.c:352 #7 0x000000010005c3c0 in runScenarioBody (c=0x100115b18) at mal_scenario.c:386 #8 0x000000010005c5ea in runScenario (c=0x100115b18) at mal_scenario.c:420 #9 0x0000000100017d35 in MSserveClient (dummy=0x100115b18) at mal_session.c:360 #10 0x00007fff846f4dcb in _pthread_start () #11 0x00007fff846f4c8d in thread_start () The thread 5 and 6 seem to both wait on the same lock. It looks like both threads are seperate clients is this possible, ie are you running multiple mclients or so?
niels
Thread 4 (process 41456 thread 0x1203): #0 0x00007fff84745886 in msync () #1 0x0000000100c38190 in MT_msync (p=0x9ac507000, off=0, len=4391849608, mode=16384) at gdk_posix.c:800 #2 0x0000000100b0ea7c in GDKsave (nme=0x101331d78 "11/1150", ext=0x100d64104 "tail", buf=0x9ac507000, size=4391849608, mode=1) at gdk_storage.c:211 #3 0x000000010089fcc6 in HEAPsave (h=0x10a8ffdb0, nme=0x101331d78 "11/1150", ext=0x100d64104 "tail") at gdk_heap.c:296 #4 0x0000000100b10dfd in BATsave (bd=0x10a02f808) at gdk_storage.c:470 #5 0x000000010089d348 in BBPsync (cnt=250, subcommit=0x10a8fff20) at gdk_bbp.c:2905 #6 0x0000000100886ce7 in TMsubcommit (b=0xc0e057458) at gdk_tm.c:115 #7 0x0000000100c3bc28 in bm_subcommit (list=0x10130ee78, catalog=0x10130ee78, extra=0xc0e07bf38, debug=0) at gdk_logger.c:838 #8 0x0000000100c3eded in bm_commit (lg=0x109f74f08) at gdk_logger.c:1559 #9 0x0000000100c3afc3 in logger_commit (lg=0x109f74f08) at gdk_logger.c:776 #10 0x0000000100c3ca0e in logger_exit (lg=0x109f74f08) at gdk_logger.c:1053 #11 0x0000000100c3cb1d in logger_restart (lg=0x109f74f08) at gdk_logger.c:1080 #12 0x0000000109dca6fd in bl_restart () #13 0x0000000109dc0013 in store_manager () #14 0x0000000109d82a37 in mvc_logmanager () #15 0x00007fff846f4dcb in _pthread_start () #16 0x00007fff846f4c8d in thread_start () Thead 4 currently holds the lock, its busy committing data to disk. This process could take quite some time (depending on the io system). It indeed will keep your clients hanging until its done.
Niels
Thread 3 (process 41456 thread 0x1103): #0 0x00007fff84712526 in select$DARWIN_EXTSN () #1 0x00000001087f1e35 in SERVERlistenThread (Sock=0x1094ba0f8) at mal_mapi.c:152 #2 0x00007fff846f4dcb in _pthread_start () #3 0x00007fff846f4c8d in thread_start () Thread 2 (process 41456 thread 0xf03): #0 0x00007fff846cdd02 in __semwait_signal () #1 0x00007fff84735f27 in nanosleep () #2 0x0000000100c3882f in MT_sleep_ms (ms=5000) at gdk_posix.c:1774 #3 0x000000010096c45b in GDKvmtrim (limit=0x100d74868) at gdk_utils.c:1350 #4 0x00007fff846f4dcb in _pthread_start () #5 0x00007fff846f4c8d in thread_start () Thread 1 (process 41456 thread 0x10b): #0 0x00007fff846cdd02 in __semwait_signal () #1 0x00007fff84735f27 in nanosleep () #2 0x0000000100c3882f in MT_sleep_ms (ms=5000) at gdk_posix.c:1774 #3 0x0000000100002dd7 in main (argc=7, av=0x7fff5fbfee80) at mserver5.c:514
On Thu, Mar 19, 2009 at 6:01 PM, Yue Sheng <[1]yuesheng8@gmail.com> wrote:
Managed to COPY 161 files to one table and another 162 files to another table. But when I do a simple select count(*) from table2; it failed (nothing, just hangs)....
Has anyone successfully run this database on the MAC, in 64bit, with table (not database) size larger than 500million rows (10 columns)?
On Thu, Mar 19, 2009 at 8:59 AM, Yue Sheng <[2]yuesheng8@gmail.com> wrote:
Now have tried the same bulk load with COPY on the latest (v5.10.0) with single thread setting.
Failed again!
System:
OS X
MonetDB complied in 64bit
number of files to load: 322
total size of files: 15GB
Max row in file: 3million
On Wed, Mar 18, 2009 at 11:51 AM, Martin Kersten <[3]Martin.Kersten@cwi.nl> wrote:
Yue Sheng wrote:
I'm not sure how "The (parallel) load used scratch area as well" is related to the question.
If you look at the code, you will notice that there is a two phase loading process involving (possibly) multiple threads
Sorry if I'm a bit slow.
On Wed, Mar 18, 2009 at 11:25 AM, Martin Kersten <[4]Martin.Kersten@cwi.nl <mailto:[5]Martin.Kersten@cwi.nl>> wrote: Yue Sheng wrote: Sorry, if I wasn't clear on the first question: (1) we ramp up N for the first insert to claim sufficient space. Sure, understand that one. But: The claimed space got "given back" *right after* the first insert. (this is the part I don't understand.) The (parallel) load used scratch area as well Question: how does the second, third, .... inserts get the "benefit" of the ramp up that we did for the first insert? Is this a bit clearer what my question pertains? Thanks. On Wed, Mar 18, 2009 at 10:26 AM, Martin Kersten <[6]Martin.Kersten@cwi.nl <mailto:[7]Martin.Kersten@cwi.nl>
<mailto:[8]Martin.Kersten@cwi.nl <mailto:[9]Martin.Kersten@cwi.nl>>> wrote: Yue Sheng wrote: Three questions that bothers me are: (1) why we need to ramp up N to total of all line in first insert. to let the kernel claim sufficient space Reason I ask is that right after first insert, the allocation drop right down from, say 100GB to 35GB, and stays roughly there for *all* subsequent inserts. I totally do not understand this. (2) in your opinion, based on this experience, what could be the potential problem here? little to none, as the files are memory mapped, which only may cause io on some systems (3) in your opinion, would the newer version cure the problem? a system can never correctly guess what will come, especially since the source of a COPY command need not be a file but standard input, i.e. a stream. Thanks. On Tue, Mar 17, 2009 at 10:51 PM, Martin Kersten <[10]Martin.Kersten@cwi.nl <mailto:[11]Martin.Kersten@cwi.nl> <mailto:[12]Martin.Kersten@cwi.nl <mailto:[13]Martin.Kersten@cwi.nl>> <mailto:[14]Martin.Kersten@cwi.nl <mailto:[15]Martin.Kersten@cwi.nl> <mailto:[16]Martin.Kersten@cwi.nl <mailto:[17]Martin.Kersten@cwi.nl>>>> wrote: Yue Sheng wrote: Martin, It almost worked... This is what I did and what have happened: I have 322 files to insert into data base, totaling 650 million rows I divided the file list into two, then for each sub list (a) I insert first file in the list with N set to 650milllion rows, (b) all subsequent files have N set to the number of lines in *that* file once first list done, then (c) I insert first file in the second list with N set to 650million rows, (d) all subsequent files have N set to the number of lines in *that* file Then the same problem happened: it stucked at file number 316. ok. using the 650M enables MonetDB to allocate enough space and does not have to fall back on guessing. Guessing is painful, because when a file of N records has been created and it needs more, it makes a file of size 1.3xN. This leads to memory fragmentation. in your case i would have been a little mode spacious and used 700M as a start, because miscalculation of 1 gives a lot of pain. Such advice is only needed in (a) Note: This is farther then previous tries, which all stopped in the region of file number 280 +/- a few. My observation: (i) at (a), the VSIZE went up to around 46GB, then after first insert, it drops to around 36GB ok fits (ii) at (c), the VSIZE went up to around 130GB, then after first insert, it drops to around 45GB you tell the system to extend existing BATs prepare for another 650 M, which means it allocates 2*36 G, plus room for the old one gives 108GB then during processings some temporary BATs may be needed,e.g. to check integrity constraints after each file,. Then it runs out of swapspace. (iii) the "Free Memory", as reported by Activity Monitor, just before it failed at file number 316, dipped to as low as 7.5MB! yes, you are running out of swapspace on your system. This should not have happened, because the system uses mmapped files and may be an issue with the MacOS or relate to a problem we fixed recently My question: (1) why we need to ramp N up to total number of lines (it takes along time to do that), then only have it drop down to 30GB-40GB right after this might indicate that on MacOS, just like Windows, mmaped files need to be written to disk. With a disk bandwidth of 50MB/sec it still takes several minutes the first insert and stay roughly there? Does it mean we're giving back all the pre-allocation space back to the OS? Then should we set N always to total number of lines? If so, it would take much much longer to process all the files... (2) How come RSIZE never goes above 4GB? (3) Does sql log file size have some limit, that we need to tweak? no limit (4) Has anyone successfully implemented the 64bit version of MondeDB and successfully inserted more than 1billion rows? you platform may be the first, but Amherst has worked with Macs for years (5) when you say you "...The VSIZE of 44G is not too problematic, i am looking at queries letting it tumble between 20-80 GB....," What does it mean? Mine went up to as high as 135GB... explained above. regards, Martin Thanks, as always.
References
1. mailto:yuesheng8@gmail.com 2. mailto:yuesheng8@gmail.com 3. mailto:Martin.Kersten@cwi.nl 4. mailto:Martin.Kersten@cwi.nl 5. mailto:Martin.Kersten@cwi.nl 6. mailto:Martin.Kersten@cwi.nl 7. mailto:Martin.Kersten@cwi.nl 8. mailto:Martin.Kersten@cwi.nl 9. mailto:Martin.Kersten@cwi.nl 10. mailto:Martin.Kersten@cwi.nl 11. mailto:Martin.Kersten@cwi.nl 12. mailto:Martin.Kersten@cwi.nl 13. mailto:Martin.Kersten@cwi.nl 14. mailto:Martin.Kersten@cwi.nl 15. mailto:Martin.Kersten@cwi.nl 16. mailto:Martin.Kersten@cwi.nl 17. mailto:Martin.Kersten@cwi.nl
------------------------------------------------------------------------------
_______________________________________________ MonetDB-users mailing list MonetDB-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/monetdb-users
-- Niels Nes, Centre for Mathematics and Computer Science (CWI) Science Park 123, 1098 XG Amsterdam, The Netherlands room C0.02, phone ++31 20 592-4098, fax ++31 20 592-4312 url: http://www.cwi.nl/~niels e-mail: Niels.Nes@cwi.nl
------------------------------------------------------------------------------ _______________________________________________ MonetDB-users mailing list MonetDB-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/monetdb-users
-- Niels Nes, Centre for Mathematics and Computer Science (CWI) Science Park 123, 1098 XG Amsterdam, The Netherlands room C0.02, phone ++31 20 592-4098, fax ++31 20 592-4312 url: http://www.cwi.nl/~niels e-mail: Niels.Nes@cwi.nl
participants (2)
-
Niels Nes -
Yue Sheng