[08:49:30] <gehel>	 dcausse: Security would like to do a pen-test on WDQS. Would you be available to do a scoping meeting on Oct 6 4:30 UTC (6:30 CEST)?
[08:49:58] <dcausse>	 gehel: sure
[08:50:34] <gehel>	 I'll cc you on that email. I'll try to be there as well (I'm interested to see how we do pentesting here), but the timing isn't great for me.
[08:51:07] <dcausse>	 is WCQS part of it? if yes it'd be nice to have Erik for the oauth system
[08:51:19] <gehel>	 I don't think so, but let me ask.
[08:52:08] <gehel>	 I suspect that this meeting is just to define the scope and the risks. We probably don't need to have too many people at that point.
[08:52:22] <dcausse>	 oh ok
[08:52:23] <gehel>	 The findings are going to be more interesting for Erik (or others).
[09:08:10] <gehel>	 ejoseph: around for our 1:1?
[09:53:39] <gehel>	 Lunch 
[10:45:57] <pfischer>	 lunch + errands
[13:09:31] <inflatador>	 <o/
[13:20:15] <dcausse>	 o/
[15:20:20] <gehel>	 ERC meeting conflicting with our unmeeting. I'll skip the unmeeting and see you all next week! Enjoy the weekend!
[18:35:48] * ebernhardson wishes alertmanager accepted syntax such as 1e9, or 100mb instead of long numbers where i count the zeros
[18:47:41] * ebernhardson would also accept underscores, like 500_000_000
[18:52:34] <ebernhardson>	 overall i'm just generally unimpressed with promtool tests, they work but are tedious. Turns out if you misspell the expected alert it says expected: foo got: [], implying no alert was given, instead of saying an alert with a different name was given.  And then when the name matches but the alert varies it doesn't show a diff, but instead a multi-line json blob for each with different key
[18:52:36] <ebernhardson>	 ordering and it's up to you to find the diff
[19:02:13] <ebernhardson>	 the context is we are seeing circuit breaking increments from the elasticsearch exporters, they seem to correlate well with young gc pool having < 500M available so trying to re-tune the alerts. Would be nice if they could restart instances if it hasn't alerted in the last N hours (days?) and only alert if the restart isn't applicable
[19:06:33] <ebernhardson>	 i suppose it would have to ensure a green cluster before restarting...maybe only auto restart if the instance has been "up" for a certain time period rather than time since last alert, or some such
[19:14:09] <ebernhardson>	 i restarted a couple instances, all in the small clusters (psi/omega) that were in the dangerous territory for now
[20:02:14] <ebernhardson>	 phan has the best error messages: PhanTypeMissingReturn Method \CirrusSearch\Maintenance\UpdateOneSearchIndexConfig::fatalError is declared to return never in phpdoc but has no return value
[20:02:34] <ebernhardson>	 because it things never is a class name i suppose :S
[20:02:38] <ebernhardson>	 s/things/thinks/