| Time |
Nick |
Message |
| 00:02 |
|
blobmarley joined #evergreen |
| 00:22 |
|
blobmarley joined #evergreen |
| 00:28 |
|
blobmarley joined #evergreen |
| 00:49 |
|
blobmarley joined #evergreen |
| 00:53 |
|
blobmarley4 joined #evergreen |
| 01:56 |
|
blobmarley joined #evergreen |
| 02:02 |
|
blobmarley joined #evergreen |
| 02:22 |
|
blobmarley4 joined #evergreen |
| 02:25 |
|
blobmarley joined #evergreen |
| 07:15 |
|
collum joined #evergreen |
| 08:41 |
|
mmorgan joined #evergreen |
| 08:53 |
|
mantis joined #evergreen |
| 09:02 |
|
dguarrac joined #evergreen |
| 09:37 |
|
smayo joined #evergreen |
| 10:26 |
|
Christineb joined #evergreen |
| 10:27 |
|
sandbergja joined #evergreen |
| 10:38 |
|
ianskelskey joined #evergreen |
| 11:36 |
jeffdavis |
On a 3.14 server, I'm finding that asset.record_has_holdable_copy is super slow (3-4 seconds to return results, vs <0.01s on a 3.11 system). |
| 11:37 |
jeffdavis |
It looks like the cause is an index in asset.copy that was introduced in 3.12: |
| 11:37 |
jeffdavis |
CREATE INDEX cp_extant_by_circ_lib_idx ON asset.copy(circ_lib) WHERE deleted = FALSE OR deleted IS FALSE; |
| 11:37 |
jeffdavis |
If I modify this index to remove the deleted=false conditions, I get normal performance again. |
| 11:38 |
jeffdavis |
Is anyone else seeing this on 3.12+? Any reason to have those deleted=false conditions on this index? |
| 11:38 |
Bmagic |
is jit disabled on the DB? |
| 11:43 |
jeffdavis |
hmm, no it isn't, that's weird |
| 11:49 |
|
StomproJ joined #evergreen |
| 11:52 |
|
jihpringle joined #evergreen |
| 11:58 |
jeffdavis |
I still see slowness after `set jit_above_cost to -1;` followed by vacuum analyze |
| 11:59 |
Bmagic |
What version of PG? Do you have an example query I can use to compare on my machines |
| 12:00 |
jeffdavis |
PG 14. Query is just `SELECT * FROM asset.record_has_holdable_copy(:rec_id)` |
| 12:07 |
|
jihpringle joined #evergreen |
| 12:18 |
csharp_ |
gmcharlt++ |
| 12:27 |
Bmagic |
jeffdavis: would you mind running that in explain analyze, and posting the results in https://explain.depesz.com/ |
| 12:28 |
jeffdavis |
explain won't show details for a function, but here's the output for a query that is functionally equivalent (and equally slow): https://explain.depesz.com/s/8xly |
| 12:28 |
jeffdavis |
and for comparison, the same query on a 3.11 server that uses the same db server but isn't slow: https://explain.depesz.com/s/bUX8 |
| 12:29 |
jeffdavis |
(full query is in the Query tab at the depesz link) |
| 12:30 |
Bmagic |
I'm getting sub 2ms execution times |
| 12:31 |
Bmagic |
EG 3.13 PG 15 |
| 12:35 |
|
ianskelskey joined #evergreen |
| 12:39 |
Bmagic |
jeffdavis: I think* we have a case of a bad plan here. Perhaps you can change the values for effective_cache_size, wal_buffers, shared_buffers, max_connections. Restart and test. See if you can get it to make a better plan when provided different values |
| 12:40 |
Bmagic |
The time killer is the way it decided to join on the function actor.org_unit_descendants. Which is a different plan compared to the "fast" version |
| 12:53 |
|
collum joined #evergreen |
| 13:46 |
csharp_ |
Planning Time: 4.406 ms |
| 13:46 |
csharp_ |
Execution Time: 0.243 ms |
| 13:46 |
csharp_ |
PG15 EG 3.14.3 |
| 13:58 |
|
jvwoolf joined #evergreen |
| 14:11 |
pinesol |
News from commits: LP1991801 Refresh after PO activation <https://git.evergreen-ils.org/?p=Evergreen.git;a=commitdiff;h=643b408c2060595632d5bef3d0d088cbf8425a72> |
| 14:17 |
|
jihpringle joined #evergreen |
| 14:47 |
|
jihpringle joined #evergreen |
| 15:27 |
|
mantis left #evergreen |
| 16:41 |
pinesol |
News from commits: LP2042988 Change icon to 'Expand All' in Acq PO <https://git.evergreen-ils.org/?p=Evergreen.git;a=commitdiff;h=5d0cd8869466023e24909f68cc3c4380db0811b2> |
| 16:46 |
|
jihpringle joined #evergreen |
| 16:55 |
berick |
claiming 1452 |
| 17:03 |
|
mmorgan left #evergreen |
| 17:11 |
pinesol |
News from commits: LP#2093899: lazy-load bucket lists in staff catalog search results <https://git.evergreen-ils.org/?p=Evergreen.git;a=commitdiff;h=320975bec224c99cc0377e2f4401193af50961de> |
| 17:41 |
pinesol |
News from commits: LP2009725 Stamping Upgrade: Recent Patron Holds <https://git.evergreen-ils.org/?p=Evergreen.git;a=commitdiff;h=b155532ae54ab71da0847cc328c79731087b221e> |
| 17:41 |
pinesol |
News from commits: LP#2009725 Place Holds for Recent Patrons <https://git.evergreen-ils.org/?p=Evergreen.git;a=commitdiff;h=d6a4eae56b34d1355a0a11246578f67ad4737168> |
| 17:58 |
|
jvwoolf joined #evergreen |
| 18:06 |
|
blobmarley joined #evergreen |
| 18:34 |
|
blobmarley joined #evergreen |
| 19:22 |
|
jihpringle joined #evergreen |
