| Time |
Nick |
Message |
| 01:41 |
|
dbs joined #evergreen |
| 01:46 |
|
book` joined #evergreen |
| 01:48 |
|
eby joined #evergreen |
| 01:48 |
|
jeffdavis joined #evergreen |
| 06:01 |
pinesol |
News from qatests: Failed Running perl live tests <http://testing.evergreen-ils.org/~live//archive/2022-05/2022-05-16_04:00:02/test.49.html> |
| 06:58 |
|
rjackson_isl_hom joined #evergreen |
| 07:06 |
|
mantis joined #evergreen |
| 07:38 |
|
rjackson_isl_hom joined #evergreen |
| 07:47 |
|
collum joined #evergreen |
| 07:48 |
|
RFrasur joined #evergreen |
| 08:37 |
|
mmorgan joined #evergreen |
| 09:04 |
|
jvwoolf joined #evergreen |
| 09:25 |
|
Dyrcona joined #evergreen |
| 09:40 |
|
collum joined #evergreen |
| 12:16 |
|
jihpringle joined #evergreen |
| 12:34 |
|
collum joined #evergreen |
| 12:38 |
|
collum joined #evergreen |
| 13:04 |
jeffdavis |
My SIP service has mysteriously stopped working, with "Unable to connect to ILS" errors in the logs. (Maybe it heard me complaining about using SIP for auth?) |
| 13:15 |
Dyrcona |
jeffdavis: Are the OpenSRF services running? |
| 13:17 |
jeffdavis |
yep, everything else is running fine on the server |
| 13:20 |
jeffdavis |
false alarm, SIP client account had expired :( |
| 13:21 |
Dyrcona |
Ah, yeah. |
| 13:52 |
|
scottangel joined #evergreen |
| 14:15 |
Dyrcona |
While investigating a Pg process crash (just some Pg processes crashed, not the postmaster), I've noticed what appears to be a pattern with some hold pull list queries: they seem to use 276 MB of temp space. |
| 14:19 |
Dyrcona |
Hmm. I guess it isn't as consistent as it looked at first glance. There are some using much less space. |
| 14:22 |
Dyrcona |
Oh. It looks like there are 2 log entries for the same query/PID, one with a higher temp space and the other lower. If wonder if that's because it grabs the extra space later or it ends up using less after some initial allocation? |
| 14:27 |
JBoyer |
I think Pg makes periodic status updates on queries and their temp space if you have temp space logging enabled. Not sure of the schedule offhand though. |
| 14:27 |
JBoyer |
(And possibly only when it changes?) |
| 14:36 |
Dyrcona |
JBoyer: I have it set to 0 to log all temp space use. Since every pull list query starts at 276MB, give or take, then there's an entry withing the same minute that's around 114KB, I assume it's grabbing the big number then dropping to the lower one. |
| 14:37 |
Dyrcona |
Hmmm. I wonder if it's copy a table, then deleting it once it's done with it? |
| 15:35 |
jeff |
postgresql docs for log_temp_files state that "a log entry is emitted for each temporary file when it is deleted" |
| 15:35 |
jeff |
I'd hazard a guess that the larger temp file is a hash-join, and the smaller is related to sorting. |
| 15:36 |
jeff |
explain analyze might give you something to correlate with your log entries. |
| 15:38 |
Dyrcona |
jeff: Yeah. Something like that. I've got one in the logs that was about 1GB yesterday. We have run out of temp space before, but that's not the problem this time. |
| 15:38 |
jeff |
whether it's hash-join or sort related (or something else, I think there are other options), I read that as one part of the query is using 276MB more than work_mem and the other is using 114KB more than work_mem. |
| 15:39 |
Dyrcona |
We had out of memory errors Saturday morning, and a bunch of these queries were running, plus some others. |
| 15:40 |
Dyrcona |
It's interesting how consistent the temp space use is for the ahopl query. |
| 15:42 |
Dyrcona |
I doubt that increasing work mem would help. :) |
| 15:42 |
|
jihpringle joined #evergreen |
| 15:48 |
jeff |
well, in theory adjusting work_mem by X would reduce the size of those temp files by X... if you have a default hash_mem_multiplier. |
| 16:01 |
Dyrcona |
jeff: I was thinking more about the out of memory situation. |
| 16:07 |
* jeff |
nods |
| 16:20 |
|
jvwoolf left #evergreen |
| 17:15 |
|
mmorgan left #evergreen |
| 18:01 |
pinesol |
News from qatests: Failed Running perl live tests <http://testing.evergreen-ils.org/~live//archive/2022-05/2022-05-16_16:00:03/test.49.html> |
| 18:15 |
|
pinesol joined #evergreen |
| 18:16 |
|
eglogbot joined #evergreen |
| 18:16 |
|
Topic for #evergreen is now Welcome to #evergreen (https://evergreen-ils.org). This channel is publicly logged. |
