This issue tracker has been migrated to GitHub, and is currently read-only.
For more information, see the GitHub FAQs in the Python's Developer Guide.

Author db3l
Recipients db3l, paul.moore, pdgoins, steve.dower, tim.golden, vstinner, zach.ware
Date 2018-05-07.05:39:57
SpamBayes Score -1.0
Marked as misclassified Yes
Message-id <1525671598.67.0.682650639539.issue33355@psf.upfronthosting.co.za>
In-reply-to
Content
Disk space seems unlikely as a reason - the local disk is 60GB on each buildbot, of which at least 50GB+ is generally free.  So that shouldn't be a problem (the WinXP/Win7 buildbots operate with far less, like 5-10GB).

I had also considered that perhaps something went wrong with the temp folder setting during the reboot so it ended up on the system disk (network) rather than the local disk, but it looks fine, and I see build and temporary files all created on the local disk.

I believe some of the worst overheads for Spectre patches were found in I/O intensive workloads.  Interestingly, neither buildbot processor reports PCID support, which might make it worse (though I don't know for sure if Azure exposes those flags to guests).

Of course, it could just be some other change in the new physical machines in some way.  There have been no other guest changes for a while (last was a build tool change in January).

Whatever it is, it certainly seems correlated to the maintenance, and appears to be consistent after that point.

-- David
History
Date User Action Args
2018-05-07 05:39:58db3lsetrecipients: + db3l, paul.moore, vstinner, tim.golden, pdgoins, zach.ware, steve.dower
2018-05-07 05:39:58db3lsetmessageid: <1525671598.67.0.682650639539.issue33355@psf.upfronthosting.co.za>
2018-05-07 05:39:58db3llinkissue33355 messages
2018-05-07 05:39:57db3lcreate