*** dyki has joined #ingres | 00:01 | |
*** dyki has quit IRC | 00:53 | |
* grantc is away: wanders off | 01:07 | |
*** Dejan has joined #ingres | 01:10 | |
Dejan | hi guys | 01:11 |
---|---|---|
*** magnusg| has joined #ingres | 01:21 | |
*** Gerhard has joined #ingres | 02:07 | |
*** withdefault has joined #ingres | 02:17 | |
*** ChanServ sets mode: +o withdefault | 02:17 | |
*** Gerhard has quit IRC | 02:20 | |
grantc | morning | 02:32 |
grantc | Dejan: regarding your suggestion for easyingres to use lighttpd - perhaps you might want to help Cedric (aka littlesatan...) make the change | 02:33 |
Dejan | hi grantc | 02:34 |
Dejan | to be frank with you, i did not see easyingres :D | 02:34 |
Dejan | i would never have a web server on windows | 02:34 |
Dejan | but if littlesatan needs my help with regards to lighty, i would help, naturally | 02:35 |
grantc | i think you suggested that you would like to see it use lighty | 02:35 |
Dejan | i will check easyingres these days | 02:35 |
grantc | :) | 02:35 |
Dejan | i did | 02:35 |
Dejan | because lighty is way better web server than apache | 02:35 |
Dejan | (IMHO) | 02:35 |
grantc | does it support ldap groups yet? | 02:36 |
Dejan | i switched to lighty some 5-6 years ago | 02:36 |
Dejan | no idea | 02:37 |
Dejan | never used ldap | 02:37 |
Dejan | :D | 02:37 |
grantc | the last time i tried it did not ... about 6 months ago | 02:38 |
Dejan | i see it can | 02:38 |
Dejan | http://wiki.netbsd.se/Lighttpd_on_NetBSD | 02:38 |
Dejan | via mod_auth | 02:39 |
grantc | it does ldap but not groups | 02:39 |
Dejan | wait, you are asking for groups... | 02:39 |
Dejan | yeah | 02:39 |
Dejan | http://redmine.lighttpd.net/issues/show/1817 | 02:40 |
Dejan | this? | 02:40 |
grantc | yup | 02:41 |
Dejan | some 5-6 yeards ago i had a home server | 02:43 |
Dejan | a very nice SGI VisualWorkstation 320 machine | 02:44 |
Dejan | adapted to be a small server | 02:44 |
Dejan | had 2 PIII CPUs | 02:44 |
Dejan | and 128MB RAM :D | 02:44 |
Dejan | amazing machine | 02:44 |
Dejan | (well, SGI pwns) | 02:44 |
Dejan | back to the story... | 02:44 |
Dejan | ... i had apache on it | 02:44 |
Dejan | and most of the time apache used whole RAM | 02:45 |
Dejan | and 90%+ CPU | 02:45 |
Dejan | i wanted to buy more RAM for that machine | 02:45 |
Dejan | but memory modules were crazily expensive | 02:45 |
Dejan | so, i decided one day to switch everything to Lighttpd | 02:45 |
Dejan | and see how it goes | 02:45 |
Dejan | i backed up everything, installed Lighty | 02:46 |
Dejan | and since then light used less than 50% CPU, and around 30 Mb of RAM | 02:46 |
Dejan | Apache was not badly configured | 02:46 |
Dejan | it was configured specially for small server like that | 02:47 |
Dejan | lighty was not configured at all | 02:47 |
Dejan | i was better straight away | 02:47 |
Dejan | same was with another, production website | 02:47 |
Dejan | that i administrated for some time | 02:47 |
Dejan | poor apache just could not cope with high load | 02:48 |
grantc | have you looked at cherokee? i have heard that's supposed to be quite good | 02:52 |
Dejan | nope | 03:32 |
Dejan | as i am very happy with lighty, for years | 03:32 |
atrofast | Dejan: how does lighttpd work with mod_wsgi? | 05:14 |
atrofast | Or does it at all? :) | 05:15 |
Dejan | never tried | 05:16 |
Dejan | i love fcgi | 05:16 |
atrofast | I guess not: http://redmine.lighttpd.net/issues/1523 | 05:27 |
*** cthibert has joined #ingres | 05:32 | |
*** mull has joined #ingres | 05:49 | |
Dejan | guys, can i somehow request tm to give me line number where the error appeared? | 05:56 |
*** rossand has joined #ingres | 06:03 | |
*** ChanServ sets mode: +o rossand | 06:03 | |
Dejan | atrofast: i suppose Python developers would like to have it... | 06:15 |
Dejan | i never liked Python :/ | 06:15 |
Dejan | i am sure mod_wsgi is there somewhere | 06:15 |
Dejan | already | 06:15 |
atrofast | Not liked Python??? I love Python :) | 06:16 |
Dejan | yeah, i never liked it | 06:16 |
Dejan | and i still do not like it | 06:16 |
Dejan | if people had Lua when Python emerged | 06:17 |
Dejan | Python would, IMHO, just be ignored | 06:17 |
Dejan | at least in areas where Lua is clearly a better choice | 06:18 |
*** magnusg| has left #ingres | 06:59 | |
Dejan | guys, is there a limit in number of IDENTITY columns? | 07:36 |
pboro | in a single table? | 07:36 |
Dejan | nope, in a database | 07:36 |
Dejan | i normally have 1 identity column per table | 07:36 |
Dejan | 10.0.0-106 crashes all the time after one CREATE TABLE statement | 07:37 |
Dejan | when i execute that particular statement, i get no error | 07:37 |
Dejan | error log says "Unexpected ULM error detected by QSF." | 07:38 |
Dejan | pretty weird | 07:39 |
pboro | congrats, you found another bug :) | 07:39 |
Dejan | yeah, but i cannot isolate this bug | 07:39 |
Dejan | and i cannot send the SQL because it is the structure of our central company database | 07:40 |
Dejan | approx. 10k lines of SQL c ode | 07:40 |
grantc | can you not replace the column names with something anonymouse? | 07:46 |
*** dyki has joined #ingres | 07:58 | |
*** cthibert has quit IRC | 08:00 | |
*** cthibert has joined #ingres | 08:01 | |
toumi01 | Dejan: is there a E_UL* message in the log when you get the E_QS000C_ULM_ERROR? | 08:05 |
Dejan | toumi01: funny thing, now i get different error | 08:09 |
Dejan | :))))) | 08:09 |
toumi01 | I'm glad you have a sense of humor about ULM. ;) | 08:09 |
Dejan | well, there is nothing funny about it | 08:10 |
Dejan | this makes my head boil | 08:10 |
toumi01 | I know :-( | 08:10 |
Dejan | because it always hangs at the same table | 08:10 |
Dejan | when i isolate table, and run that statement | 08:10 |
Dejan | it works without problems | 08:11 |
toumi01 | an example even with renames to tab1 tab2 ... col1 col2 col3 to hide business secrets would be most helpful | 08:11 |
toumi01 | I wonder if support ever wrote a script to automate that sort of thing | 08:12 |
Dejan | toumi01: I have ~ 580 tables | 08:12 |
Dejan | in this SQL file | 08:12 |
Dejan | :( | 08:12 |
toumi01 | probably the whole work load is needed to create the error :( | 08:12 |
Dejan | yeah | 08:12 |
Dejan | my guess is that IDENTITY fields are bugged | 08:13 |
Dejan | still | 08:13 |
Dejan | why i think so? | 08:13 |
Dejan | - because problem disappears when i switch to use sequences | 08:13 |
toumi01 | that _is_ strong evidence | 08:13 |
Dejan | but i cannot proof anything | 08:13 |
Dejan | that is just my guess, based from experience | 08:14 |
Dejan | :/ | 08:14 |
Dejan | we have wasted like 2 days | 08:14 |
Dejan | trying to find out what is wrong | 08:15 |
Dejan | some 2 weeks ago | 08:15 |
toumi01 | we need a version of the ingres parser that spits out the statements with arbitrary identifiers to hide the real ones | 08:15 |
Dejan | we thought we had some weird special character in our SQL | 08:15 |
toumi01 | these subtle bugs lead to lots of good theories about the root cause | 08:15 |
toumi01 | but many of those theories will turn out to be false | 08:16 |
Dejan | one thing i am 100% is that something is bugged with IDENTITY fields | 08:16 |
Dejan | no doubt about it | 08:16 |
Dejan | toumi01: i think i will write a proggy for obfuscating SQL code | 08:34 |
Dejan | when i go back home | 08:34 |
Dejan | well, not really obfuscating | 08:34 |
toumi01 | that would be a very valuable tool! | 08:35 |
toumi01 | anonymizing? in terms of ids? | 08:35 |
Dejan | yes | 08:38 |
Dejan | guys, this is the error i get most of time: http://codepad.org/x794QbDj | 08:44 |
Dejan | this "scs_sequencer" is suspicious there | 08:45 |
toumi01 | scs_sequencer is the core dbms engine state machine | 08:47 |
toumi01 | it does not have anything particular to do with sequences (in spite of the name) | 08:47 |
toumi01 | but the back trace is very interesting | 08:47 |
toumi01 | I'm right now working on a cache_dynamic race condition that seems to involve qso_destroy | 08:48 |
* toumi01 wonders if QSF is the source of all evil in the world | 08:48 | |
pboro | :D | 08:50 |
pboro | toumi01, ahh, there's still race(s) in cache_dynamic... | 08:50 |
toumi01 | pboro: yes, the code is very dynamic ;) | 08:51 |
pboro | I'm currently hitting a race in new LOB handling of JDBC, reported it over a month ago and it's still unresolved :/ | 08:51 |
toumi01 | is this the one with readlock=nolock? | 08:52 |
pboro | I have a sense that Ingres code base is pretty sensitive to changes since there's always so much new bugs introduced with new features or something | 08:52 |
toumi01 | the code base is not inherently sensitive (good compartmentalization) | 08:53 |
toumi01 | but there are times when "mistakes were made", as the saying goes | 08:53 |
toumi01 | it's hard to create QA workloads that are representative of real-world workloads :( | 08:54 |
pboro | hmm nope, we use readlock = shared | 08:54 |
pboro | this is 138181 | 08:55 |
*** DerMeister has joined #ingres | 08:59 | |
*** grantc has quit IRC | 09:13 | |
*** Dejan has quit IRC | 10:04 | |
*** M6eis6ter6 has joined #ingres | 10:45 | |
*** withdefault1 has joined #ingres | 10:56 | |
*** DerMeister has quit IRC | 10:59 | |
*** withdefault has quit IRC | 11:13 | |
*** cthibert has left #ingres | 11:14 | |
*** rossand1 has joined #ingres | 11:42 | |
*** rossand has quit IRC | 11:56 | |
*** M6eis6ter6 has quit IRC | 13:06 | |
*** DerMeister has joined #ingres | 13:07 | |
*** DerMeister has quit IRC | 14:03 | |
*** toumi01 has left #ingres | 14:28 | |
*** zxiiro has joined #ingres | 14:57 | |
*** ChanServ sets mode: +v zxiiro | 14:57 | |
*** DarylM has quit IRC | 15:26 | |
*** DarylM has joined #ingres | 15:26 | |
*** mull has quit IRC | 15:32 | |
*** DarylM has quit IRC | 15:32 | |
*** stephen1 has joined #ingres | 15:38 | |
*** stephen2 has joined #ingres | 15:47 | |
*** toumi01 has joined #ingres | 15:47 | |
*** stephenb has quit IRC | 15:57 | |
*** withdefault has joined #ingres | 15:57 | |
*** ChanServ sets mode: +o withdefault | 15:57 | |
*** withdefault2 has joined #ingres | 15:59 | |
*** stephen1 has quit IRC | 16:05 | |
*** withdefault1 has quit IRC | 16:15 | |
*** withdefault has quit IRC | 16:16 | |
*** toumi01 has left #ingres | 16:17 | |
*** zxiiro has quit IRC | 16:31 | |
*** rossand1 has quit IRC | 18:18 | |
*** mull has joined #ingres | 19:21 | |
*** mull has quit IRC | 20:27 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!