[04:55:18] <pfischer>	 ryankemper: Awesome, thank you! That’s great news. 🎉
[13:28:23] <ebernhardson>	 \o
[13:28:59] <inflatador>	 <o/
[13:34:44] <ebernhardson>	 hmm, cindy is starting to have generic failures again :( 
[13:35:16] <ebernhardson>	 usually that means it's timing out or otherwise failing in a one of the setup routines
[13:55:02] <pfischer>	 o/
[13:55:31] <pfischer>	 ebernhardson: you mean in the context of https://gerrit.wikimedia.org/r/c/mediawiki/extensions/CirrusSearch/+/1067928?checksRunsSelected=test ?
[13:56:00] <ebernhardson>	 pfischer: yea basically, there were 2 tests it -1'd last night, at least one the day before as well
[13:56:26] <ebernhardson>	 all with the same error, usually when it fails a bunch of things it's because a cached setup routine they all depend on failed
[13:57:29] <pfischer>	 Would that be sth. to investigate during the DPE pairing session?
[13:58:19] <ebernhardson>	 we could look a bit, but i'd have to do a school run in ~30 min.  Usually it's a big guessing game to pin down what exactly is failing. 
[13:58:52] <ebernhardson>	 well not entirely guessing, there are logs and ways to run subsets of the tests
[13:59:27] <pfischer>	 Okay, I’d be interested to get some insights
[14:00:17] <ebernhardson>	 sure, joining
[15:02:52] <gehel>	 Wednesday meeting starting in https://meet.google.com/eki-rafx-cxi
[15:59:40] <inflatador>	 errand, back in ~40
[18:09:34] <inflatador>	 sorry, been back...but taking lunch and will be back in ~40
[18:29:34] <ebernhardson>	 random story time: Trey and i looked into why the search metrics dashboard had a weird change on aug 12 and 13. The summary looks to be someone used a botnet to visit every page in every wikipedia over those two days
[18:31:32] <ebernhardson>	 which changed the denominator of the metrics, which gives the funky shape
[19:37:38] <tax>	 How exactly does Blazegraph slicing work? If I send a query with a slice directive to two different endpoints, will I get the same content, or are different data.jnl files going to produce different results?
[19:38:33] <ebernhardson>	 tax: i'm sad to say that it's unlikely anyone here knows. David might have ideas, but he's on vacation
[19:39:29] <tax>	 I think the safest thing would be for me to execute multi-part queries on the same instance. I use slices to get every Wikidata item with a DOI across four queries (ten million per slice)
[19:39:31] <ebernhardson>	 tax: from a test and see perspective,  there are many servers backing the wdqs site, if you get different answers from issuing the same query that could be suggestive.  But also blazegraph isn't clustered, so while we try to keep the servers the same there arent strong guarantees
[19:40:37] <ebernhardson>	 tax: i'm not super famliar with your use case, but it sounds like maybe LDF would do it?
[19:40:48] <tax>	 I've observed. I run independent runUpdate.sh processes for each of my servers. And they should be very similar but also subtly different. And Blazegraph doesn't do ACID semantics does it
[19:42:31] <ebernhardson>	 tax basically, https://query.wikidata.org/bigdata/ldf?subject=&predicate=p%3AP356&object=
[19:43:07] <ebernhardson>	 tax: indeed blazegraph doesn't have transactions or ACID in the way sql does
[19:43:17] <ebernhardson>	 iirc LDF runs against a single server to give consistency to the results at least
[19:44:05] <ebernhardson>	 (at least, our ldf endpoint)