Fri, Apr 3
Given April 22 is now going to be a WMF holiday, will the talk be cancelled or postponed?
@Krinkle do you know when this started happening?
Thu, Apr 2
Was fixed already.
I find it odd that the "
Wed, Apr 1
Tue, Mar 31
Mon, Mar 30
The fix has been merged.
Wed, Mar 25
Wrote up a more detailed checklist for making a Scap release. https://gerrit.wikimedia.org/r/c/mediawiki/tools/scap/+/583319
Mon, Mar 23
I'm thinking the following now:
Thu, Mar 19
Ack, I'm working on an acceptance test suite for applying patches, and will be basing the implementation on the scap patch plugin.
Wed, Mar 18
Same as T247075?
I would recommend against dropping the use of -o pipefail. It's one of the safeguards against the many pitfalls in writing shell scripts. It'd be better to invoke bash instead, in my opinion.
Wed, Mar 11
This change has been merged and the resulting CI post-merge failure has been addressed. Closing task as finished.
I'm working on a test environment in which to run the test suite. It turns out that the scap Debian package is built in jessie, but the scap test suite doesn't actually run on jessie. I'm going to see about fixing that.
Apparently there is a scap plugin for this already. I'll be looking at that and adding instructions for using it to the train deployment page.
Mar 6 2020
http://git.liw.fi/wmf-ssh-config/ is the test suite, http://git.liw.fi/liw-dot-files/tree/ssh/ is the config. Works for me now. Might be generalizable for others, if there's interest. Test suite requires https://subplot.liw.fi/ (a hobby project of mine).
Mar 5 2020
Group2 has been running 1.35.0-wmf.22 for almost an hour and there's been no explosion of these errors, so I'm removing it as a train blocker.
After discussion on IRC it seems likely this probably, maybe, won't happen more often if we promote train to group2. Leaving as a train blocker, will promote go group2 and remove as train blocker later if things go well.
Marking as UBN due to risk to train.
Mar 4 2020
I note that something like this just happend in group1 for 1.35.0-wmf.22:
Unblocking train for this task.
Rolled forward to group1 again, after waiting for a wikidata edit spike from being over.
Rolles back due to T246898
Group1 deployed just now.
It seems the oldest Python3 we have in production is 3.4, so we need to keep that as the minimum baseline when porting.
Cool, and thanks. That SWAT deploy seems to have happened after the logstash messages I saw. Closing task.
Mar 3 2020
Group0 deployed earlier today, and looking OK so far.
Started cutting the branch for this week's train.