| Time |
Nick |
Message |
| 07:07 |
|
JBoyer joined #evergreen |
| 07:43 |
|
collum joined #evergreen |
| 07:49 |
|
kworstell-isl joined #evergreen |
| 08:06 |
|
BDorsey joined #evergreen |
| 08:17 |
|
rfrasur joined #evergreen |
| 08:36 |
|
mmorgan joined #evergreen |
| 08:44 |
|
Stompro joined #evergreen |
| 09:09 |
|
Dyrcona joined #evergreen |
| 09:31 |
|
mantis1 joined #evergreen |
| 09:42 |
|
dguarrac joined #evergreen |
| 09:42 |
|
dmoore_doppel joined #evergreen |
| 09:52 |
|
dmoore_beta joined #evergreen |
| 09:59 |
|
sandbergja joined #evergreen |
| 10:11 |
|
collum joined #evergreen |
| 10:39 |
|
collum joined #evergreen |
| 10:51 |
|
BrianK joined #evergreen |
| 11:05 |
pinesol |
News from commits: LP#2029160 - Documentation - marc_export example sql command update <https://git.evergreen-ils.org/?p=Evergreen.git;a=commitdiff;h=74bed760422c352b0798c04db03163999154bb60> |
| 13:00 |
|
collum joined #evergreen |
| 13:12 |
|
Stompro joined #evergreen |
| 13:42 |
|
kworstell-isl joined #evergreen |
| 14:14 |
jeffdavis |
I seem to be running into issues with Shibboleth (for SSO) in a multi-server environment. It looks like the same issue as bug 1992024 but we've already applied the fix for that. Not seeing the same problem on a test server. |
| 14:14 |
pinesol |
Launchpad bug 1999823 in OpenSRF "duplicate for #1992024 Name collision causes apache gateway modules to fail when mod_shib is installed" [Medium,Confirmed] https://launchpad.net/bugs/1999823 |
| 14:17 |
jeffdavis |
Specifically the same behavior as 1992024 - with mod_shib enabled I get a 502 error when I try to load the reporter. We have the fix for 1999823 which resolved the same issue in a test environment. |
| 14:48 |
JBoyer |
jeffdavis, when you mention multi-server, do you mean multiple apache servers? And if so, what StorageService backend are you using? |
| 14:51 |
jeffdavis |
By multi-server I mean four full-stack EG servers (Evergreen + Apache + nginx etc) sitting behind a load balancer. |
| 14:52 |
|
mantis1 left #evergreen |
| 15:02 |
jeffdavis |
("full stack" probably isn't the right term but we're not running Apache and EG on separate hosts or anything) |
| 15:10 |
JBoyer |
I mean, that's how I use the term, right or no. :) the real Q was "> 1 apache" which was yes. I've been distracted looking for paste sites since it looks like we trashed our local service since the last time I needed to use one. |
| 15:12 |
* jeff |
defaults to gist.github.com |
| 15:13 |
JBoyer |
That's probably what I'll do from now on after seeing pastebin.com's wall of ads. But, jeffdavis, I suspect your various shibd's aren't sharing session info unless you've done something similar to this: https://pastebin.com/0Rpi5FNC |
| 15:14 |
JBoyer |
Note that memcached server doesn't *have* to be the one(s) used by Evergreen, so long as all shibd's are using the same one(s). |
| 15:16 |
jeff |
my other default/reflex is to turn links like that into https://pastebin.com/0Rpi5FNC |
| 15:16 |
jeff |
er. |
| 15:16 |
jeff |
mispaste. irony. |
| 15:16 |
JBoyer |
Oh, and I forgot that the SessionCache stanza needs a StorageService="mc-ctx" attribute, and both ReplayCache and ArtifactMap need a StorageService="mc" attribute so they know to use the memcached one. |
| 15:16 |
jeff |
inserting raw/ in the mix like this is handy https://pastebin.com/raw/0Rpi5FNC |
| 15:17 |
JBoyer |
ah, self-ad-block. jeff++ |
| 15:18 |
jeffdavis |
uuugggggghhhhhhhhh |
| 15:18 |
jeffdavis |
but also JBoyer++ |
| 15:18 |
* Dyrcona |
uses and ad blocker and paid for a lifetime pastebin subscription. (I figured it was worth supporting a tool that I use.) |
| 15:19 |
* Dyrcona |
knows nothing about shibboleth integration. |
| 15:21 |
JBoyer |
jeffdavis, here's a gist that's better and includes those other tags. |
| 15:21 |
JBoyer |
https://gist.github.com/HitScan/71f043e043e04a8eb58f48b29c4b8528 |
| 15:21 |
* JBoyer |
just bookmarks gist.github.com for next time. |
| 15:22 |
Dyrcona |
I use gist.github for some things. Did you know you can clone the gist and treat it like a git repo? That's a neat feature. |
| 15:22 |
jeffdavis |
thank you! |
| 15:23 |
* Dyrcona |
tries to get the volume just right on the Def Leppard to drown out the backround noise, but also not go deaf. |
| 15:23 |
JBoyer |
That is handy. I also forget that they can have comment threads which is also handy |
| 15:23 |
* JBoyer |
reaches for a thesaurus for synonyms of 'handy' |
| 15:24 |
Dyrcona |
utlitaritarian |
| 15:25 |
JBoyer |
Oops, maybe I won't use <> for placeholders in xml as that is misleading. |
| 15:25 |
* Dyrcona |
is trying to come up with a SQL to remove a couple of org units and keeps running into trigger issues. I swear this worked two months ago. I think someone experimented with these in produciton in the mean time. |
| 15:27 |
Dyrcona |
Jobyer: [] might work. |
| 15:28 |
JBoyer |
I just went with PLACEHOLDER, but yeah, () and [] are fine substitutes |
| 15:29 |
jeffdavis |
Not once in multiple years of poking at Shibboleth integration have I come across the idea that I need to specify a storage service like memcached. We're not even trying to use it for auth yet, just seeing how EG works when mod_shib is enabled. It is disturbing that some shared session backend is already needed at this stage. |
| 15:30 |
Dyrcona |
Oh. I thought you wanted something where you could do substitution of multiple values, like [name], [age], etc. |
| 15:30 |
jeffdavis |
(not complaining about the EG integration here, just ... shib seems incredibly brittle and makes apache brittle too) |
| 15:30 |
JBoyer |
I tend to go with @AUTOMAKE@ style if I want a program to do it. |
| 15:30 |
Dyrcona |
Software is brittle in general. The hardware ain't so tough, either. |
| 15:31 |
jeffdavis |
some software is more brittle than others though |
| 15:31 |
JBoyer |
jeffdavis, well, you shouldn't need to do that just to get things installed. I thought you turned it on and things weren't working. If you're getting 5XX errors just having shibd running that isn't a good sign/ |
| 15:31 |
Dyrcona |
I've been using string.Template in Python, so ${var} is common. If I have something really complicated, I might break out TT2. |
| 15:32 |
Dyrcona |
jeffdavis++ True dat! |
| 15:34 |
jeffdavis |
To be clear, Shibboleth is installed and shibd is running on EG servers where Apache runs but we're not using it yet for EG auth. Enabling mod_shib in this environment causes 5XX errors connecting to the gateway (again, even when EG isn't actually trying to use it). |
| 15:35 |
JBoyer |
Then my pastes won't help (outside of the usual Old Magic of turning it off and back on again) |
| 15:35 |
JBoyer |
:-/ |
| 15:36 |
jeffdavis |
heh, well maybe it's solving my next problem instead of my current one ;) |
| 15:45 |
JBoyer |
At least it's not falling over dead like the libssh thing, you may be able to track it down by just cranking up the logging until it tells you why it's mad. |
| 16:09 |
|
BDorsey joined #evergreen |
| 16:30 |
Dyrcona |
Um.... Same data, and I get a "new" conflict: ERROR: update or delete on table "org_unit" violates foreign key constraint "hold_request_selection_ou_fkey" on table "hold_request" DETAIL: Key (id)=(2) is still referenced from table "hold_request".\ |
| 16:30 |
Dyrcona |
That one didn't happen on my other copy of this data. |
| 16:31 |
* Dyrcona |
tries a third copy. (I've got 4 readily available.) |
| 16:32 |
Dyrcona |
I wonder if maybe that constraint is missing or not yet available in one of the databases. The four databases are on 3 different Evergreen releases. |
| 16:34 |
Dyrcona |
Yeah, that's it! |
| 16:34 |
Dyrcona |
The constraint is newer than 3.10.3. |
| 17:09 |
|
mmorgan left #evergreen |
| 18:00 |
jeffdavis |
Looks like my shib issue is intermittent. I pulled one of the servers from the load balancer and tried refreshing on /eg/staff/reporter/legacy/main repeatedly. 50% of the time the page loaded correctly; the other 50% of the time console logs show a 502 Bad Gateway error |
| 18:00 |
jeffdavis |
when I get the 502 error, nginx error logs show this message: upstream prematurely closed connection while reading response header from upstream, client: <ip>, server: , request: "GET /reports/fm_IDL.xml HTTP/1.1", upstream: "https://127.0.0.1:7443/reports/fm_IDL.xml" |
| 18:01 |
jeffdavis |
Very weird that it's so intermittent. Some kind of race condition? |
| 18:30 |
jeffdavis |
Interestingly https://localhost/gateway?format=json&service=open-ils.actor&method=opensrf.open-ils.system.ils_version does NOT give me a 502 error, intermittent or otherwise - it consistently loads for me even when the reporter doesn't. |
| 19:16 |
jeff |
Perhaps a problem with mod_idlchunk. Perhaps not the exact same issue as bug 1999823, but something to look at next. |
| 19:16 |
pinesol |
Launchpad bug 1999823 in OpenSRF "Name collision causes apache gateway modules to fail when mod_shib is installed" [Medium,Confirmed] https://launchpad.net/bugs/1999823 |
| 19:16 |
jeff |
mod_idlchunk because that module is in play with your /reports/fm_IDL.xml request. |
| 19:18 |
jeffdavis |
worth looking into for sure, thanks for the idea |
| 19:24 |
jeff |
agreed that it being 50/50 is a bit interesting. |
| 19:25 |
jeff |
does the request for /reports/fm_IDL.xml fail reliably? does it fail only when going through the proxy? |
| 21:47 |
|
stompro joined #evergreen |
