- users-list - monetdb.org

[MonetDB-users] Mserver hangs on quit(); on RedHat 8.0 (psyche)
by Stefan.Manegold＠cwi.nl 15 Dec '04

15 Dec '04

Hi again, with the latest MonetDB 4.5.0 on RedHat 8.0 (psyche), Jan (F.;-) experiences a "consistently" hang of the Mserver, once he tries to quit it. See below for details. It did work with older versions of MonetDB, but we hanven't locate more precisely, yet, since which version/code change the hang occurs. My question is, has anybody else ever experienced similar problems? I now vaguely recall, that there might have been some related issue, but I don't remember the details, yet --- could someone who does refresh my mind? Thank you very much in advance! Stefan Forwarded message: > From flokstra(a)cs.utwente.nl Wed Dec 15 09:30:02 2004 > X-Sieve: CMU Sieve 2.2 > Content-Type: text/plain; > charset="us-ascii" > From: Jan Flokstra <flokstra(a)cs.utwente.nl> > Organization: University Twente Enschede > To: Stefan.Manegold(a)cwi.nl (Stefan Manegold) > Subject: Re: Monet problem > Date: Wed, 15 Dec 2004 09:27:45 +0100 > User-Agent: KMail/1.4.3 > References: <20041213155606.9656241044(a)draco.ins.cwi.nl> > In-Reply-To: <20041213155606.9656241044(a)draco.ins.cwi.nl> > Cc: "Peter Boncz" <P.Boncz(a)cwi.nl> > MIME-Version: 1.0 > Old-Content-Transfer-Encoding: quoted-printable > Message-Id: <200412150927.47729.flokstra(a)cs.utwente.nl> > X-Spam-Score: -4.9 () BAYES_00 > X-Scanned-By: milter-spamc/0.25.320 (styx.cwi.nl [192.16.191.21]); Wed, 15 Dec 2004 09:28:11 +0100 > X-Scanned-By: MIMEDefang 2.44 > X-Spam-Flag: NO > X-Spam-Status: NO, hits=0.00 required=5.00 > X-Spam-Level: > X-Spam-Report: Content analysis details: (0.0 points, 5.0 required) > Tests (points, rulename, description): > ____ > X-Virus-Scanned: by cwi.nl (CWI/$Revision: 1.25 $) > Content-Transfer-Encoding: 8bit > > Hi Stefan, > > I am using Redhat Linux, version (uname -a) > > Linux psyche 2.4.18-14 #1 Wed Sep 4 13:35:50 EDT 2002 i686 i686 i386 GNU/Linux > > and the default compilers on this version. I made an strace of the Mserver > and got the following output (just the tail) > It looks it never returns from the rt_sigsuspend() syscall and this is > indeed very strange! > > Regards, Jan. > > > brk(0x8175000) = 0x8175000 > brk(0x8177000) = 0x8177000 > brk(0x8178000) = 0x8178000 > brk(0x8179000) = 0x8179000 > brk(0x817b000) = 0x817b000 > brk(0x817d000) = 0x817d000 > brk(0x817f000) = 0x817f000 > brk(0x8180000) = 0x8180000 > brk(0x8181000) = 0x8181000 > brk(0x8182000) = 0x8182000 > brk(0x8183000) = 0x8183000 > brk(0x8185000) = 0x8185000 > brk(0x8186000) = 0x8186000 > close(7) = 0 > open("/usr/local/flokstra/lib/MonetDB", > O_RDONLY|O_NONBLOCK|O_LARGEFILE|O_DIRECTORY) = 7 > fstat64(7, {st_mode=S_IFDIR|0755, st_size=8192, ...}) = 0 > fcntl64(7, F_SETFD, FD_CLOEXEC) = 0 > lseek(7, 0, SEEK_SET) = 0 > getdents64(0x7, 0x80a1b90, 0x1000, 0x4006c4eb) = 4096 > open("/usr/local/flokstra/lib/MonetDB/trans.mil", O_RDONLY|O_LARGEFILE) = 8 > fstat64(8, {st_mode=S_IFREG|0755, st_size=70, ...}) = 0 > fstat64(8, {st_mode=S_IFREG|0755, st_size=70, ...}) = 0 > mmap2(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = > 0x4045f000 > read(8, "# This module is generated using"..., 4096) = 70 > read(8, "", 4096) = 0 > close(8) = 0 > munmap(0x4045f000, 4096) = 0 > close(7) = 0 > open("/usr/local/flokstra/lib/MonetDB", > O_RDONLY|O_NONBLOCK|O_LARGEFILE|O_DIRECTORY) = 7 > fstat64(7, {st_mode=S_IFDIR|0755, st_size=8192, ...}) = 0 > fcntl64(7, F_SETFD, FD_CLOEXEC) = 0 > lseek(7, 0, SEEK_SET) = 0 > getdents64(0x7, 0x80a1b90, 0x1000, 0x400709c6) = 4096 > getdents64(0x7, 0x80a1b90, 0x1000, 0x400709c6) = 2504 > open("/usr/local/flokstra/lib/MonetDB/lib_trans.so", O_RDONLY) = 8 > read(8, "\177ELF\1\1\1\0\0\0\0\0\0\0\0\0\3\0\3\0\1\0\0\0\314\16"..., 1024) = > 1024 > fstat64(8, {st_mode=S_IFREG|0755, st_size=78034, ...}) = 0 > old_mmap(NULL, 16136, PROT_READ|PROT_EXEC, MAP_PRIVATE, 8, 0) = 0x419f7000 > mprotect(0x419fa000, 3848, PROT_NONE) = 0 > old_mmap(0x419fa000, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED, 8, > 0x2000) = 0x419fa000 > close(8) = 0 > write(1, "MonetDB>", 8MonetDB>) = 8 > fstat64(0, {st_mode=S_IFCHR|0600, st_rdev=makedev(136, 1), ...}) = 0 > mmap2(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = > 0x4045f000 > read(0, quit(); > "quit();\n", 1024) = 8 > write(6, "\200\20G@\2\0\0\0\0\0\0\0\222\253\0@\304\363\377\277\300"..., 148) = > 148 > rt_sigprocmask(SIG_SETMASK, NULL, [RTMIN], 8) = 0 > rt_sigsuspend([] > > > -- | Dr. Stefan Manegold | mailto:Stefan.Manegold@cwi.nl | | CWI, P.O.Box 94079 | http://www.cwi.nl/~manegold/ | | 1090 GB Amsterdam | Tel.: +31 (20) 592-4212 | | The Netherlands | Fax : +31 (20) 592-4312 |

1 0

[MonetDB-users] "pseudo" nested BATs
by Stefan.Manegold＠cwi.nl 15 Dec '04

15 Dec '04

Hi MonetDB developers and users, sorry for this "blunt" posting, but I thought, the problem, respectively the solution/recommendation that will hopefully be initiated by this question, might be interesting for some more of you: Triggered by a question from Agustin, I noticed that I (believe it or not ;-) apparently lack some knowledge of / experiences with MonetDB/MIL: Consider you have a variable "a" holding a (possible transient) BAT, and another variable "b" holding a second BAT, which happens to be a view of the first one. Both BATs haven't been baptized ("rename()"), yet; hence they still hold there default names: var a := new(void,int).seqbase(0(a)0).insert(nil,1).insert(nil,2).access(BAT_READ); var b := reverse(a); print(a); #-----------------# # h tmp_29 # name # void int # type #-----------------# [ 0@0, 1 ] [ 1@0, 2 ] print(b); #-----------------# # t tmp_29 # name # int void # type #-----------------# [ 1, 0@0 ] [ 2, 1@0 ] Then, these BATs should be gathered in a "container"-BAT. Since (persistent) nested BATs have disappeared (for good reasons!), we do this by gathering the BAT's name in the container: var c := new(void,str).seqbase(10@0); c.insert(nil,bbpname(a)).insert(nil,bbpname(b)); print(c); #-------------------------# # h tmp_28 # name # void str # type #-------------------------# [ 10@0, "tmp_29" ] [ 11@0, "tmp_29" ] Since the second BAT is only a view of the first one, they share not only the same memory/storage, but also the same (bbp-)name; hence, they are not distinguishable by (only) their (bbp-)names, and thus the container c actually contains the first BAT twice iso each BAT once: print(bat(fetch(c,0))); #-----------------# # h tmp_29 # name # void int # type #-----------------# [ 0@0, 1 ] [ 1@0, 2 ] print(bat(fetch(c,1))); #-----------------# # h tmp_29 # name # void int # type #-----------------# [ 0@0, 1 ] [ 1@0, 2 ] print(bbpname(a)); [ "tmp_29" ] print(bbpname(b)); [ "tmp_29" ] Obviously, the problem can be solved/avoided by baptizing the two BATs, and hence making them unique "individuals": rename(a,"BAT_a"); rename(b,"BAT_b"); print(bbpname(a)); [ "BAT_a" ] print(bbpname(b)); [ "BAT_b" ] c.insert(nil,bbpname(a)).insert(nil,bbpname(b)); print(c); #-------------------------# # h tmp_28 # name # void str # type #-------------------------# [ 10@0, "tmp_29" ] [ 11@0, "tmp_29" ] [ 12@0, "BAT_a" ] [ 13@0, "BAT_b" ] print(bat(fetch(c,0))); !ERROR: interpret_params: print(param 1): invalid BAT. print(bat(fetch(c,1))); !ERROR: interpret_params: print(param 1): invalid BAT. print(bat(fetch(c,2))); #-----------------# # h BAT_b # name # void int # type #-----------------# [ 0@0, 1 ] [ 1@0, 2 ] print(bat(fetch(c,3))); #-----------------# # t BAT_b # name # int void # type #-----------------# [ 1, 0@0 ] [ 2, 1@0 ] My question is now: - Is this the only solution, or am I just not aware of another command iso bbpname() that would return unique "default" name for non-renamed view? Of course, given the fact that only MonetDB itself has control over the default "tmp_*" names, it is always recommendable to assign names that you can control yourself, whenever you use a BAT's name as its only reference... Stefan -- | Dr. Stefan Manegold | mailto:Stefan.Manegold@cwi.nl | | CWI, P.O.Box 94079 | http://www.cwi.nl/~manegold/ | | 1090 GB Amsterdam | Tel.: +31 (20) 592-4212 | | The Netherlands | Fax : +31 (20) 592-4312 |

2 1

[MonetDB-users] Where is the user documentation in the Windows install?
by Tolkin, Steve 06 Oct '04

06 Oct '04

Dear Monet users and developers, I installed Monet on Windows XP using MonetDB-Installer-4.4.0.zip But there is no user documentation! I took the default directory "c:\program files\CWI\MonetDB" and looked in and below it -- but did not see any documentation. I want to know the basics, e.g. how to create a table in SQL, populate it, and query it. OI am guessing that a statement ends with a semicolon, but do not know even this. I am guessing that it accepts standard SQL, but that s not clear either. I am able to run the Server, and the Client, and the Mknife tool. I tried a few ways in the Mknife tool, but without success. For example in the SQL client tool I entered sql>create table t (i integer); And got back the message: # 6 # querytype I assume this is some kind of error. Here is some more of my attempts to do something. Why are the values for token number much higher than the number of tokens in my statement? sql>create tabel t (c char(1)) more>/ more>; MAPI = monetdb@localhost:45123 QUERY = create tabel t (c char(1)) ERROR = !ERROR parse error at token (258) in statement: create tabel sql>create table t (c char(1)); MAPI = monetdb@localhost:45123 QUERY = create table t (c char(1)); ERROR = !current transaction is aborted (please ROLLBACK) sql>rollback more>create table t (c char(1)); MAPI = monetdb@localhost:45123 QUERY = rollback ERROR = !ERROR parse error at token (372) in statement: rollback !create sql>rollback; # 7 # querytype sql>rollback more>/ more>; MAPI = monetdb@localhost:45123 QUERY = rollback ERROR = !ERROR parse error at token (47) in statement: rollback !/ sql>rollback more>; # 7 # querytype sql>ROLLBACK; # 7 # querytype sql> I then tried looking for the documentation in MonetDB-4.4.0.tar.gz But this explains how to compile the system, not how to use it. I think there may be user documentation in the *.rpm files intended for Linux, but I do think I can use those on Windows. Hopefully helpfully yours, Steve -- Steve Tolkin Steve . Tolkin at FMR dot COM 617-563-0516 Fidelity Investments 82 Devonshire St. V4D Boston MA 02109 There is nothing so practical as a good theory. Comments are by me, not Fidelity Investments, its subsidiaries or affiliates.

2 1