| Time |
Nick |
Message |
| 00:07 |
|
dkyle joined #evergreen |
| 00:09 |
|
_bott_ joined #evergreen |
| 00:42 |
|
JBoyer_alt joined #evergreen |
| 02:52 |
|
kipd joined #evergreen |
| 02:54 |
|
Glen joined #evergreen |
| 03:01 |
|
ejk joined #evergreen |
| 05:00 |
pinesol |
News from qatests: Testing Success <http://testing.evergreen-ils.org/~live> |
| 07:03 |
|
sandbergja joined #evergreen |
| 07:05 |
|
JBoyer joined #evergreen |
| 07:06 |
|
rjackson_isl joined #evergreen |
| 07:27 |
|
Dyrcona joined #evergreen |
| 07:52 |
|
agoben joined #evergreen |
| 08:36 |
|
mmorgan joined #evergreen |
| 08:54 |
|
sandbergja joined #evergreen |
| 09:09 |
|
stephengwills joined #evergreen |
| 09:54 |
|
Dyrcona joined #evergreen |
| 09:55 |
Dyrcona |
So, we just upgraded to 3.2.4 from 3.0.13 and we're getting "Loading.... [missing image icon]Page Loading" on the catalog search in the web staff client. |
| 09:56 |
Dyrcona |
Anyone else had this happen? |
| 09:56 |
Dyrcona |
Some sites are reporting that search doesn't load/finish, but I've not seen that personally. |
| 10:02 |
Dyrcona |
So, it looks like we're somehow missing the throbber image. |
| 10:06 |
JBoyer |
Do you have a separate static media server or anything like that? |
| 10:07 |
Dyrcona |
JBoyer: No, but I've figure id out. |
| 10:07 |
Dyrcona |
s/id/it/ |
| 10:07 |
Dyrcona |
There are supposed to be two copies of the image, apparently, and we only have 1 on production. |
| 10:08 |
JBoyer |
++ |
| 10:08 |
Dyrcona |
/openils/var/web/images/progressbar_green.gif |
| 10:08 |
Dyrcona |
/openils/var/web/opac/images/progressbar_green.gif |
| 10:08 |
Dyrcona |
We're missing the first of those. |
| 10:09 |
JBoyer |
I didn't realize they were both the same. |
| 10:10 |
Dyrcona |
And, they're neither progress bars, nor green. It's a grey, circular throbber. |
| 10:12 |
Dyrcona |
I suspect our custom image branch may have deleted one. I'll check. |
| 10:16 |
Dyrcona |
Ah ha! I just realized I forgot to start ruby webrick. |
| 10:16 |
Dyrcona |
That has nothing to do with the missing image.... I realize that came from out of nowhere. |
| 10:18 |
JBoyer |
I forgot you were still using that there. I am a fan of the new non-ruby stuff. |
| 10:20 |
JBoyer |
(Mostly because of the no-ruby parts, but not having a thing to start is also pretty nice.) |
| 10:20 |
Dyrcona |
Yeah, it's not all up to me. There's some history with being the first to upgrade to a new release here, so some staff are rather conservative about switching to new things. |
| 10:21 |
Dyrcona |
I think it was on my list, I just missed it. |
| 10:22 |
Dyrcona |
Nope. Got omitted from my upgrade-specific steps, but it's on my list of what to do when starting and stopping the servers. |
| 10:26 |
Dyrcona |
Back to the image, it is my custom images branch.... |
| 10:27 |
Dyrcona |
"Remove a pile of unused images and move 1 to opac/images." from February 2017, though how it's not missing on training or wasn't missing before this upgrade, I dunno. |
| 10:27 |
Dyrcona |
It might not have been used before this upgrade, though. |
| 10:32 |
Dyrcona |
Looks like that's the only "unused" image that is still in master. I guess the others were ours. |
| 10:41 |
JBoyer |
Yeah, I don't think that loading page was there in 3.0, I don't remember if it was added in 3.1 or 2. |
| 10:44 |
|
collum joined #evergreen |
| 10:58 |
|
yboston joined #evergreen |
| 10:58 |
Dyrcona |
Well, easy matter to put the image back, just git checkout the file from master. |
| 10:58 |
Dyrcona |
git++ |
| 11:03 |
Bmagic |
well, it's been too many years. I am running up against the search path unaccent error again. The function unaccent(text) clearly exists in the public schema. show search path; shows that public is the first search path. but the logs complain about the function not existing |
| 11:05 |
Dyrcona |
Bmagic: Is the database extension installed? That uses the database unaccent database extension. If that' s not installed, it won't exist. |
| 11:05 |
jeff |
what's the log message? |
| 11:05 |
Bmagic |
I bet that's it |
| 11:06 |
Bmagic |
ERROR: function unaccent(text) does not exist at character 28 |
| 11:06 |
jeff |
also there are search paths that need to be explicit for both referencing the function and for referencing the dictionary. |
| 11:07 |
Bmagic |
CREATE EXTENSION unaccent; - answer extension "unaccent" already exists |
| 11:08 |
jeff |
that looks like an unqualified reference to unaccent(), what's the context? |
| 11:08 |
JBoyer |
You're restoring a db? |
| 11:08 |
jeff |
is this a pg_restore or normal usage? |
| 11:08 |
Bmagic |
the db is analyzing actor.usr |
| 11:09 |
jeff |
bug 1671150 has a bit of info that you might want to review, if you haven't already |
| 11:09 |
Bmagic |
this DB was just freshly synced from another database. Same version 9.5 |
| 11:09 |
pinesol |
Launchpad bug 1671150 in Evergreen 3.0 "Unqualified references in evergreen.unaccent_and_squash lead to index creation failures with pg_restore" [Medium,Fix released] https://launchpad.net/bugs/1671150 |
| 11:09 |
jeff |
"the db is analyzing" = VACUUM ANALYZE? |
| 11:09 |
jeff |
"freshly synced from another database" using what method? |
| 11:09 |
Bmagic |
auto analyze is what the logs say |
| 11:09 |
Bmagic |
rsync |
| 11:10 |
jeff |
okay, autovacuum. got it. |
| 11:10 |
Bmagic |
rsync while postgres was stopped on both |
| 11:10 |
jeff |
can you share output of \df+ evergreen.unaccent_and_squash |
| 11:12 |
jeff |
and \d+ actor.usr |
| 11:12 |
pastebot |
"Bmagic" at 64.57.241.14 pasted "\df+ evergreen.unaccent_and_squash" (10 lines) at http://paste.evergreen-ils.org/10847 |
| 11:12 |
|
ejk joined #evergreen |
| 11:13 |
jeff |
okay, that looks outdated. what version of Evergreen? |
| 11:13 |
jeff |
should look closer to RETURN evergreen.lowercase(public.unaccent('public.unaccent', regexp_replace(arg, '[\s[:punct:]]','','g'))); |
| 11:13 |
|
kipd joined #evergreen |
| 11:13 |
Bmagic |
yeah, I was about to say that. Maybe I just need to run the upgrade scripts |
| 11:13 |
Bmagic |
it's 2.11 |
| 11:13 |
|
Glen joined #evergreen |
| 11:14 |
Dyrcona |
I don't know what the #postgresql community says, but rsync is not how I would copy a Pg database. I'd pg_dump and then pg_restore. |
| 11:14 |
jeff |
ah. that didn't get the fix. you can probably apply 1083 but you should verify. the fix was only backported as far as 2.12.8 according to the bug above. |
| 11:14 |
Bmagic |
jeff++ |
| 11:14 |
Dyrcona |
jeff++ |
| 11:15 |
Bmagic |
Drycona: I've been using rsync since I started working with Evergreen. The method that was passed down to me from hopkinsju. When transitioning from one version of postgres to another, I use dump/restore |
| 11:15 |
|
sandbergja joined #evergreen |
| 11:16 |
Dyrcona |
OK. Just saying that's not how I would do it. |
| 11:16 |
Bmagic |
It seems the rsync method would take less time |
| 11:22 |
jeff |
if you hit this during a pg_restore, it would have been because the restored evergreen.unaccent_and_squash function had an assumption/dependency on the search_path being set to something other than what pg_restore sets it to. |
| 11:23 |
Dyrcona |
It would take less time, but I'd suspicious of it missing links that can be created by the Postgres db upgrade commmand. |
| 11:23 |
jeff |
if you're encountering the search path related errors you pasted above during autovacuum (and not pg_restore) operation, you might lack the database setting which sets search path. that might cause other issues. |
| 11:24 |
jeff |
(though I would like to continue to eliminate said search path issues / requirements as we identify them) |
| 11:24 |
jeff |
\drds in the psql shell will show database-level settings. |
| 11:25 |
jeff |
i believe the current evergreen-recommended database setting for search path is: search_path=evergreen, public, pg_catalog |
| 11:26 |
Dyrcona |
Well evergreen is automatically added if your database user is named evergreen. |
| 11:31 |
berick |
Bmagic: just saw this guy on cpan, thought of you... https://metacpan.org/author/OALDERS |
| 11:39 |
jeff |
oh, actually the autoanalyze search_path issue is probably just something that happens in any post-CVE-2018-1058 postgres version, because part of the fix for that was to empty the search path when autovacuum runs: https://git.postgresql.org/gitweb/?p=postgresql.git;a=commitdiff;h=91f3ffc5249eff99c311fb27e7b29a44d9c62be1 |
| 11:39 |
pinesol |
A flaw was found in the way Postgresql allowed a user to modify the behavior of a query for other users. An attacker with a user account could use this flaw to execute code with the permissions of superuser in the database. Versions 9.3 through 10 are affected. (http://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2018-1058) |
| 11:42 |
jeff |
Bmagic: comment 3 on bug 1671150 is relevant to you -- I'm pretty sure (or at least, past me is pretty sure) that you shouldn't just apply that db upgrade script to 2.11. |
| 11:42 |
pinesol |
Launchpad bug 1671150 in Evergreen 3.0 "Unqualified references in evergreen.unaccent_and_squash lead to index creation failures with pg_restore" [Medium,Fix released] https://launchpad.net/bugs/1671150 |
| 12:04 |
|
aabbee joined #evergreen |
| 12:12 |
|
sandbergja joined #evergreen |
| 12:22 |
Bmagic |
berick++ |
| 12:29 |
|
khuckins joined #evergreen |
| 12:31 |
|
Dyrcona joined #evergreen |
| 12:32 |
|
sandbergja joined #evergreen |
| 12:39 |
Bmagic |
jeff: I'm a bit confused about "not running the upgrade script" - The DB in this example was upgraded to 2.11 a few years ago (pre git commit for said bug) - Therefore, it seems that the git fix should be included in the up coming sql scripts that I need to make the transition to 3.1 |
| 12:41 |
Dyrcona |
Bmagic: I believe jeff made the comment about not running the upgrade script before he knew what release the database was on. |
| 12:41 |
Dyrcona |
Though I haven't checked the logs for the last hour, yet. :) |
| 12:43 |
Bmagic |
I think I understand: don't run this fix on 2.11 directly. But if I am running the full set of upgrade scripts to take the DB to 3.1 then it's fine? |
| 12:43 |
Bmagic |
which should cover the issue brought up in comment #3 about bug 1613341 diffs |
| 12:43 |
pinesol |
Launchpad bug 1613341 in Evergreen "Patron name search - apostrophes" [Wishlist,Fix released] https://launchpad.net/bugs/1613341 |
| 12:44 |
jeff |
Bmagic: I was suggesting that you not apply only database upgrade script 1083 to your 2.11 system as a backported fix. Past me stated in the bug that the backport to 2.11 would need to take into consideration changes that had been made in parallel. If you're upgrading away from 2.11 then that's probably not a concern. |
| 12:44 |
Bmagic |
cool! I understand |
| 12:44 |
jeff |
Bmagic: yes, i think you've got it! |
| 12:45 |
Bmagic |
jeff++ Dyrcona++ |
| 12:45 |
Dyrcona |
Yeah, what jeff said. :) |
| 12:45 |
* Dyrcona |
just back from the logs. |
| 13:07 |
|
rjackson_isl joined #evergreen |
| 14:10 |
|
sandbergja joined #evergreen |
| 14:24 |
|
sandbergja joined #evergreen |
| 14:49 |
|
sandbergja joined #evergreen |
| 15:12 |
|
khuckins joined #evergreen |
| 15:23 |
Dyrcona |
@band search Bad Gateway |
| 15:23 |
pinesol |
Dyrcona: No bands found matching 'Bad Gateway' |
| 15:24 |
Dyrcona |
@band add Bad Gateway |
| 15:24 |
pinesol |
Dyrcona: Band 'Bad Gateway' added to list |
| 15:26 |
berick |
@coffee [ana [band]] |
| 15:26 |
* pinesol |
brews and pours a cup of Kenya, and sends it sliding down the bar to Ha Ha! CC |
| 15:26 |
berick |
kinda surprised that worked |
| 15:26 |
berick |
at least, I think it did ;) |
| 15:27 |
berick |
@tea [coffee [ana [band [who]]]] |
| 15:27 |
* pinesol |
brews and pours a pot of Honey Black Tea, and sends it sliding down the bar to brews and pours a cup of Nectar, and sends it sliding down the bar to Ulcerative daring (http://ratetea.com/tea/health-and-tea/honey-black-tea/7529/) |
| 15:28 |
berick |
well, alright |
| 15:29 |
jeff |
neat. websocketd died. |
| 15:30 |
jeff |
...or maybe not |
| 15:31 |
jeff |
nope. false alarm. |
| 15:31 |
* berick |
exhales |
| 15:32 |
jeff |
berick: i know, right? :-) |
| 15:34 |
jeff |
3.1 test system, i've gotten a client in a state where it's trying to delete its expired auth token, but console reports "Possibly unhandled rejection: undefined", and digging into it, the browser reports the failure as |
| 15:40 |
jeff |
er, browser reported the error as a failure to connect to the wss:// url. |
| 15:40 |
jeff |
of course, i can't get it to report that any more. |
| 15:43 |
jeff |
there we go. |
| 15:44 |
jeff |
WebSocket connection to 'wss://host.example.org/osrf-websocket-translator' failed: Unspecified reason |
| 16:06 |
jeff |
unable to reproduce in another profile. clearing cookies from dev tools got past the failure mode, then i lacked a workstation. maybe disk cleanup bit me. |
| 16:06 |
jeff |
my test offlines were still present. |
| 16:51 |
|
sandbergja joined #evergreen |
| 17:01 |
pinesol |
News from qatests: Testing Success <http://testing.evergreen-ils.org/~live> |
| 17:08 |
|
mmorgan left #evergreen |
| 20:53 |
|
Christineb joined #evergreen |
| 21:32 |
|
beanjammin joined #evergreen |
