Time |
Nick |
Message |
08:25 |
|
mmorgan joined #evergreen |
08:48 |
mmorgan |
Good morning #evergreen! |
08:48 |
mmorgan |
@coffee [someone] |
08:48 |
* pinesol |
brews and pours a cup of Panama Hacienda La Esmeralda Gesha, and sends it sliding down the bar to dbs |
08:48 |
mmorgan |
@tea [someone] |
08:48 |
* pinesol |
brews and pours a pot of Golden Orchid, and sends it sliding down the bar to eby (http://ratetea.com/tea/whispering-pines/golden-orchid/7244/) |
08:49 |
mmorgan |
Not sure who's around today, but does anyone know of issues with sms messages getting to T-Mobile? |
09:43 |
|
jvwoolf joined #evergreen |
09:53 |
|
Dyrcona joined #evergreen |
10:02 |
jvwoolf |
Reingest with --delay_symspell took over 11 hours, and I got an error back saying " temporary file size exceeds temp_file_limit (20971520kB)" when it tried to run the symspell_dictionary_full_reify function |
10:04 |
Dyrcona |
jvwoolf: Not surprising. I want to add that a full ingest on our database would take a couple of days prior to the symspell code. I've not been able to complete a full ingest with symspell enabled 'cause it fills the database drive. |
10:04 |
Dyrcona |
So, we've disabled symspell. |
10:05 |
Dyrcona |
BTW, it also filled the disk on a server with multiple terabytes of space to spare. |
10:08 |
Dyrcona |
jvwoolf: You could try truncating any relevant tables and running the steps to populate symspell from scratch. That might be faster and have a higher chance of success. |
10:13 |
jvwoolf |
Dyrcona: I might start with a fresh copy of the production DB and upgrade it again. We'd be starting from a blank slate for the production upgrade anyway. |
10:15 |
jvwoolf |
Just to clarify though, is it better to do the reindexing steps for symspell before or after a reindex, if we're making changes to our other indexes? |
10:15 |
Dyrcona |
In that case, I'd disable the symspell triggers during the ingest, re-nable them, then do the sympell steps. |
10:15 |
jvwoolf |
I'm talking about this this: https://evergreen-ils.org/documentation/release/RELEASE_NOTES_3_8.html#_reindexing_for_search_suggestions |
10:15 |
jvwoolf |
Sorry, I meant before or after the reingest |
10:16 |
Dyrcona |
My suggestion crossed your question, but I'd probably do the ingest first, then do the symspell steps from the release notes. |
10:16 |
jvwoolf |
Dyrcona++ |
10:22 |
Dyrcona |
jvwoolf: You might find this useful. https://pastebin.com/PBmHVJ1q |
10:24 |
jvwoolf |
Dyrcona++ |
10:25 |
jvwoolf |
I had those commands already from csharp_, but was hoping that patch would make them unnecessary |
10:26 |
jvwoolf |
Ooh |
10:26 |
jvwoolf |
Just came across this in Launchpad search: https://bugs.launchpad.net/evergreen/+bug/1998355 |
10:26 |
pinesol |
Launchpad bug 1998355 in Evergreen "Did You Mean's search.symspell_dictionary can get significantly bloated" [Medium,New] |
10:28 |
jvwoolf |
Also wondering about testing this with queued ingest, since we have some time |
10:34 |
Dyrcona |
I was testing queued ingest, but other things came up before I could finish. |
10:35 |
Dyrcona |
jvwoolf: That branch you linked needs a db upgrade script. |
11:25 |
|
mantis1 joined #evergreen |
12:04 |
|
mmorgan left #evergreen |
12:38 |
jvwoolf |
Dyrcona: Does CW/Mars have authority records? |
12:42 |
Dyrcona |
jvwoolf: Yes, we do have authority records. |
12:44 |
Dyrcona |
hmm. redundant index is redundant: "metabib_title_field_entry_index_vector_gin_idx" gin (index_vector) "metabib_title_field_entry_index_vector_idx" gin (index_vector) |
12:45 |
* Dyrcona |
looks up (again) how to determine how much space indexes take up. |
12:49 |
Dyrcona |
money.payment is 0 bytes? |
12:52 |
Dyrcona |
Well, that's interesting. One of those "redundant" indexes is 1 MB larger than the other. |
12:57 |
Dyrcona |
Hmm.. Looks like I have a few more of these redundant gin indexes, and I could save a few GB of space by dropping them. |
13:22 |
Dyrcona |
Yeah. Dropping 6 redundant GIN indexes knocked about 7GB out of the database. |
15:58 |
Dyrcona |
"If it keeps on rainin', the levee's goin' to break." |