[09:34:14] sigh... 93,0KB/s to download flink from archive.apache.org... [09:34:35] going to switch to a mirror from blubber [09:40:27] meh "Official Mirror Network Discontinued" since 2021 [09:41:39] they have a cdn now but this does not serve archived tar balls (older releases) [10:04:02] going to push the tar ball I have on stat1004 to people.wm.o and use that in the meantime I guess... [10:08:01] dcausse: if you have time for review the comm on WDQS data reload failure: https://docs.google.com/document/d/1rK4zg-7BfeHnftcB22IguCffWT_jD12QYhUtRvoY3Kw/edit# [10:08:11] looking [10:11:16] pfischer: less of your area for the technical side of it, but your feedback on the wording and tone is very much welcomed as well! [10:11:36] gehel: quick chat? [10:12:07] https://meet.google.com/gru-mxis-ubq [10:13:01] pfischer: ^ [11:06:37] lunch [11:38:50] lunch [11:50:29] hey! Yet more questions from WMDE! How (or are you?) getting request metrics for your Elasticsearch instance; are you using envoy? Is there some special switch in elsaticsearch to expose some kind of "successful/failed" or "latency" of requests? [12:10:48] gehel: I read your “Communication on data reload failures”: It’s informative and easy to understand (event without a lot of context). The section dealing with the impact is a bit vague. It only looks at the consumer side. Does this have an impact on the future of the WDQS project/product too? [12:15:23] pfischer: it does (I'm a lot less confident about the ability of Blazegraph to handle our data sizes outside of data reload now). But I wasn't sure if I should add something about it in this communication [12:15:39] Since you wonder, it probably means that I should [12:16:00] s/wonder/ask/ [13:02:59] tarrow: there are plenty of metrics... we use https://gerrit.wikimedia.org/r/plugins/gitiles/operations/puppet/+/2567dfcdf904f69dd17859161554e7344fe6fc5f/modules/prometheus/files/usr/local/bin/prometheus-wmf-elasticsearch-exporter.py to expose them to prometheus [13:03:58] for latencies between CirrusSearch <> elasticsearch we use envoy but also track some metrics from mediawiki itself [13:04:40] e.g. https://grafana.wikimedia.org/d/000000455/elasticsearch-percentiles?orgId=1&refresh=1m [13:04:54] so no special switch in elasticsearch, only tooling around it [14:15:58] Does anybody know how to build the analytics project locally? I’m trying to follow the Jenkins logs but tox runs into errors. So far I have python 3.7.16 (via pyenv), a venv with tox. Once I run tox -v inside the venv, it fails: python setup.py egg_info did not run successfully. [14:18:14] As far as I can tell, this happens during pytest: install_deps [14:30:04] pfischer: https://wikitech.wikimedia.org/wiki/Discovery/Analytics#Airflow_fixtures [14:30:30] usin the docker-registry.wikimedia.org/releng/tox-pyspark image [15:03:06] dcausse no hurry but I am up at https://meet.google.com/qve-fycn-vpw if/when you wanna work on the flink k8s stuff [15:03:31] inflatador: oh cool joining [16:04:41] Heading to puppet training for the rest of the day...but do ping me if it's urgent [17:02:33] i don't know why i'm surprised ... opened the fulltext head queries dashboard and the top 3 results are for a search term that means roughly nsfw movie in a non-english language [17:05:48] also reminder, next monday is a US holiday [17:12:06] dinner