*** grantc has joined #ingres | 00:09 | |
*** Mud has joined #ingres | 00:25 | |
*** PaulM05 has joined #ingres | 00:59 | |
*** KermitTheFragger has joined #ingres | 01:17 | |
Vroomfondle | just out of interest, can Ingres do online checkpointing (i.e. without exclusively-locking the entire database)? | 02:56 |
---|---|---|
Vroomfondle | Oracle/MySQL/etc seem to have ways of doing it | 02:57 |
PaulM05 | yes | 02:57 |
Vroomfondle | PaulM05: any idea where I should start looking for details on how to do it? | 02:57 |
Vroomfondle | we bring our system down for half an hour every day for checkpointing. I'm not a Systems Manager so it's not my decision, but I'd really like to be able to prod them at a better way of doing things... | 02:58 |
PaulM05 | check the Command Ref guide for the ckpdb command - checkpoints are online by default unless you use -l IIRC | 02:59 |
Vroomfondle | k, cheers | 02:59 |
PaulM05 | there will still be a short pause whilst it completes existing transactions - plus there are some restrictions (no DDL) - so you still want it to be a 'quiet time' if possible | 03:00 |
Vroomfondle | Ah, I see there's a section in the DBA guide | 03:01 |
PaulM05 | yep | 03:01 |
Vroomfondle | it seems to indicate that the checkpoint will wait until existing transactions are finished, but will prevent any new transactions from opening up during that time (presumably so it doesn't have to wait forever) | 03:01 |
PaulM05 | it will 'stall' new transactions - so they will begin eventually | 03:02 |
Vroomfondle | That sounds fine tbh | 03:02 |
PaulM05 | you can set a timeout for the stall if you want | 03:02 |
Vroomfondle | I shall go have a word with the more reasonable system manager, and see if the idea'll fly | 03:03 |
Vroomfondle | cheers | 03:03 |
PaulM05 | no problem | 03:03 |
Vroomfondle | erm... okay. Just checked the command his script runs and, erm, we're already doing online checkpointing. | 03:26 |
Vroomfondle | So we've been kicking users out for half an hour every night, for years, for no good reason. | 03:27 |
Vroomfondle | Oh joy. | 03:27 |
PaulM05 | depending on how long "years" is it might have made sense - long ago the default was offline I think | 03:27 |
grantc | :) | 03:27 |
grantc | perhaps you had uncommitted transactions which would cause the ckpdb to wait | 03:28 |
Vroomfondle | grantc: we commit after every query (don't get me started on that) | 03:29 |
grantc | :/ | 03:29 |
Vroomfondle | so with the exception of a few really enormous queries, transactions should never take long to finish ;) | 03:29 |
*** Mud has quit IRC | 03:44 | |
*** Mud has joined #ingres | 03:49 | |
*** Mud is now known as Guest27889 | 03:49 | |
*** Guest27889 has joined #ingres | 03:50 | |
*** Guest27889 is now known as Mud | 03:51 | |
*** PaulM05 is now known as PaulM05_lunch | 04:52 | |
*** Alex| has joined #ingres | 05:08 | |
*** ChanServ sets mode: +o Alex| | 05:08 | |
*** rossand has joined #ingres | 05:32 | |
*** ChanServ sets mode: +o rossand | 05:32 | |
*** rossand has quit IRC | 06:34 | |
*** gerhard has joined #ingres | 06:48 | |
*** Alex| has quit IRC | 06:59 | |
*** PaulM05_lunch is now known as PaulM05 | 07:38 | |
*** Deyan has joined #ingres | 07:43 | |
Deyan | hi everybody | 07:43 |
Deyan | any DBA around? | 07:44 |
*** Mud has quit IRC | 07:45 | |
*** Mud has joined #ingres | 07:49 | |
*** Mud has quit IRC | 08:02 | |
*** gerhard has quit IRC | 08:20 | |
FrankW | Any geospatial folks around? Chuck? Andrew? | 08:21 |
FrankW | I'm getting: E_US07DE CREATE or ALTER TABLE: invalid column format 'geometry' on column | 08:21 |
FrankW | 'geom'. | 08:21 |
FrankW | in response to: create table test (geom geometry)\g | 08:21 |
FrankW | Is there something I need to do to "geospatial enable" a database instance? | 08:22 |
*** Alex| has joined #ingres | 08:31 | |
*** ChanServ sets mode: +o Alex| | 08:31 | |
FrankW | Alex|: Are you Alex Trofast? Or perhaps another Alex? | 08:32 |
Alex| | another one :) | 08:34 |
FrankW | ok, thanks. | 08:34 |
Alex| | you are looking for "atrofast" | 08:34 |
FrankW | ok, here is a more general question for the room. Why does the limit directive not work when I do: | 08:36 |
FrankW | select * from iicolumns limit 1\g | 08:36 |
FrankW | If I remove the limit clause things work fine. | 08:37 |
FrankW | Is limit not standard sql? | 08:37 |
Alex| | no | 08:37 |
Alex| | limit is mysql and maybe some others... | 08:37 |
PaulM05 | SELECT FIRST 1 * FROM iicolumns | 08:37 |
FrankW | Ah, thanks. | 08:37 |
Alex| | ANSI is "fetch first n rows only offset m" | 08:37 |
FrankW | Thanks everyone - it seems i've picked up some non-standard habits from postgresql. | 08:38 |
Alex| | ansi sql is very chatty :) | 08:38 |
Alex| | I'm sure there is a "please" in the standards as well :) | 08:38 |
FrankW | lol | 08:39 |
Deyan | guys, do you use compression of your checkpoints? | 09:16 |
grantc | Deyan - I do | 09:17 |
Deyan | grantc: via cktmpl.def ? | 09:17 |
grantc | yup | 09:17 |
Deyan | grantc: on a production server? | 09:18 |
grantc | bugs.ingres.com | 09:18 |
Deyan | sorry if i ask too many questions :D | 09:18 |
grantc | http://ingres.pastebin.com/m44cf9be8 | 09:20 |
grantc | relies on the fact the gnutar does the compression at the same time | 09:20 |
grantc | however with backup/recovery you should test it works for you | 09:20 |
grantc | there is a variable you can set to specify an alternate cktmp.def file - PaulM05 will probably know what it is | 09:21 |
Deyan | i $II_CKTMPL_BCV i think | 09:24 |
PaulM05 | II_CKTMPL_FILE | 09:24 |
Deyan | II_CKTMPL_FILE | 09:24 |
Deyan | yeah, found that one too | 09:25 |
Deyan | making backups is a tedious work | 09:25 |
Deyan | :) | 09:25 |
Deyan | there are so many things one should do | 09:25 |
Deyan | i was thinking about not using compression from within ingres | 09:27 |
Deyan | but to write a script that will compress what needs to be compressed | 09:27 |
Deyan | :) however, i fear i may leave something important out :D | 09:28 |
*** Mud has joined #ingres | 09:51 | |
*** PaulM05 has quit IRC | 10:08 | |
*** grantc has quit IRC | 10:17 | |
*** KermitTheFragger has quit IRC | 10:22 | |
* Deyan needs SELECT column1,column2,... INTO OUTFILE 'path/to/file'; | 11:05 | |
Deyan | i need to get database names | 11:06 |
Deyan | from iidatabase | 11:06 |
Deyan | and do not want to sed TM's output... | 11:06 |
Deyan | i'll play with COPY...INTO | 11:08 |
*** withdefault has quit IRC | 11:14 | |
*** rossand has joined #ingres | 11:19 | |
*** ChanServ sets mode: +o rossand | 11:19 | |
*** Deyan has quit IRC | 11:50 | |
*** Alex| has quit IRC | 14:21 | |
*** Mud has quit IRC | 15:08 | |
*** zxiiro has quit IRC | 15:58 | |
*** cytrinox_ has joined #ingres | 19:20 | |
*** cytrinox has quit IRC | 19:24 | |
*** cytrinox_ is now known as cytrinox | 19:24 | |
*** rossand has quit IRC | 19:48 | |
FrankW | I'm getting: /usr/bin/ld: /wrk/home/warmerda/ingres/server/install/II/ingres/lib/iiclsadt.o: relocation R_X86_64_32 against `a local symbol' can not be used when making a shared object; recompile with -fPIC | 20:02 |
FrankW | I already added -fPIC to CCMACH64 in Jamdefs.hyb86_lnx and rebuild and reinstalled. | 20:02 |
FrankW | But I gather object files like iiclsadt.o are built as part of the install process. | 20:03 |
FrankW | Is there something special to make sure it is built with -fPIC? | 20:03 |
FrankW | Hmm, on review, it does seem to be a product of the regular build, not the install process. | 20:08 |
FrankW | Perhaps my Jamdefs.hyb86_lnx change was not effective for some reason. | 20:08 |
*** Alex| has joined #ingres | 22:36 | |
*** ChanServ sets mode: +o Alex| | 22:36 | |
*** Alex| has quit IRC | 23:05 | |
*** grantc has joined #ingres | 23:17 | |
*** Deyan has joined #ingres | 23:47 | |
Deyan | good morning gents | 23:48 |
*** Alex| has joined #ingres | 23:51 | |
*** ChanServ sets mode: +o Alex| | 23:51 | |
Deyan | morning Alex| | 23:56 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!