| Time |
Nick |
Message |
| 07:06 |
|
collum joined #evergreen |
| 07:50 |
|
sandbergja joined #evergreen |
| 07:51 |
sandbergja |
claiming 1472 |
| 08:08 |
pinesol |
News from commits: LP1927510: stamp upgrade script <https://git.evergreen-ils.org/?p=Evergreen.git;a=commitdiff;h=8e995fa82b234f4561d770e7d1cf001fda6ae0cb> |
| 08:08 |
pinesol |
News from commits: LP1927510 Coded Value Map Save Grid Settings <https://git.evergreen-ils.org/?p=Evergreen.git;a=commitdiff;h=56ea9d76098116427451c5de7a3c9fd5bfe845ca> |
| 08:39 |
|
mmorgan joined #evergreen |
| 08:57 |
|
mantis1 joined #evergreen |
| 09:05 |
|
dguarrac joined #evergreen |
| 09:11 |
|
Dyrcona joined #evergreen |
| 09:18 |
|
sandbergja joined #evergreen |
| 09:33 |
|
redavis joined #evergreen |
| 09:53 |
|
sandbergja joined #evergreen |
| 11:29 |
|
Christineb joined #evergreen |
| 12:04 |
|
jihpringle joined #evergreen |
| 15:33 |
eeevil |
berick: friday at 3:30 is def /not/ the time start a deep-parts-of-opensrf-on-redis discussion, so I'm just leaving this as a marker (and so I'm not just pouncing on you without a chance to refresh), but maybe next week if we have some time at the same time, I have some things I'd like to go over. tl;dr: either I'm crazy -- entirely 100% possible -- or the HA and LB stuff from the XMPP infra is not really there (some of it tries to be, but I don't |
| 15:33 |
eeevil |
think it /actually/ is). I believe I'm close to fixing that, but I don't want to go farther without a sanity check (or 10) from you as the redis bus author. |
| 15:34 |
eeevil |
I'll put the gist of the issues in an email to -dev if you want (probably can do that Monday) |
| 15:44 |
csharp_ |
@band add Redis Bus |
| 15:44 |
pinesol |
csharp_: Band 'Redis Bus' added to list |
| 15:46 |
csharp_ |
eeevil: (also in need of a weekend, but) my impression is that each instance of Redis/OpenSRF is isolated and the cross-domain stuff isn't available - or maybe I'm misunderstanding |
| 15:47 |
csharp_ |
all localhost all the time |
| 15:48 |
csharp_ |
(not taking a position, just sharing what I know about what I think you're talking about :-) ) |
| 15:49 |
eeevil |
csharp_: there's definitely the ability to register cross-domain ... berick and I worked that out (holy moly) in summer '22. but, it only ever sends requests to the first-registered service listener -- it doesn't rotate the list, or maintain a pointer to move forward in the list. (I've added a list rotation function (and some related code) to make that happen) |
| 15:51 |
csharp_ |
oh - cool |
| 15:51 |
csharp_ |
eeevil++ |
| 15:51 |
eeevil |
so, for now, yes. it's effectively all localhost all the time |
| 15:52 |
eeevil |
that is def true in practice today, you understand The Reality ;) |
| 15:53 |
csharp_ |
@band add [someone], [someone], and [someone] |
| 15:53 |
pinesol |
csharp_: Band 'mmorgan , Jaysal , and sleary' added to list |
| 15:54 |
csharp_ |
@band add OpenSRF ShmopenShrf |
| 15:54 |
pinesol |
csharp_: Band 'OpenSRF ShmopenShrf' added to list |
| 15:57 |
|
jihpringle joined #evergreen |
| 16:08 |
mmorgan |
Heh, that's an interesting band. |
| 16:09 |
* mmorgan |
can neither sing nor play an instrument! |
| 16:13 |
csharp_ |
mmorgan: no one can hear you sing on IRC!!! |
| 16:13 |
csharp_ |
lalalalala |
| 16:14 |
mmorgan |
That is a good thing! |
| 16:14 |
berick |
eeevil: hm, i'm not too surprised the router isn't load balancing. i could have missed that. I would expect the router to fall-back to the second-registered service if the first-registered disappeared, though. pretty sure I tested that specifically. but i'm all for load balancing. happy to ta |
| 16:15 |
berick |
lk more in email/lp/whatevs. |
| 16:15 |
berick |
@band add The OpenSRF Mother Smurf |
| 16:15 |
pinesol |
berick: Band 'The OpenSRF Mother Smurf' added to list |
| 16:22 |
eeevil |
yeah, it'll do the right thing if index 0 in service instance list fails (and if next one fails, etc), but it doesn't rotate. I have code for that (and polished up osrfList with Pack and Rotate). the other thing is there's no delivery failure detection at all (that's why I was beating on the STREAM drum). I've mostly got that figured out, too. the last thing is "bus reset" ... if it's rebooting the universe on service restart, that's no good. I / |
| 16:22 |
eeevil |
think/ that's what it's doing. but, I'm still just starting to dig into that part. last is ACL restrictions -- my hope is to allow routerless mode with /at least/ public/private separation. but that's currently a stretch goal ATM. and then with that, for the "bunch of standalone full servers" crowd (no cross-reg ever), an automatic routerless mode to speed things up in that simple(r) case |
| 16:23 |
berick |
eeevil: *nod* and i wouldn't worry about routerless mode anymore, I don't think that's a requirement anymore |
| 16:25 |
eeevil |
well, if it ends up be the trival change that I expect, we can have it for free and skip one redis key bounce per request, and the router resources, small as they may be. but if it's nontrivial I'll move on |
| 16:26 |
eeevil |
but I have deeper questions ... later :) |
| 16:27 |
eeevil |
thanks for (likely) confirming the top line on service instances, though! that provides some confidence that I'm not missing something obvious |
| 16:55 |
|
mmorgan left #evergreen |
| 18:40 |
|
jihpringle joined #evergreen |
| 18:56 |
|
jihpringle joined #evergreen |
