Find out why syncing descriptors from collector2.tp.o did not time out
On August 4, collector.tp.o attempted to fetch remote files from collector2.tp.o during its normal sync run. However, this process stalled, because of an unknown problem on collector2.tp.o. The effect was that neither of the two downloaded consensuses and we missed almost an entire day of consensuses. (We got them back from friendly relay operators in the end, but that should not become the new way of feeding the archive with data.)
From the log:
2018-08-04 17:37:17,556 INFO o.t.m.c.c.CollecTorMain:75 Starting sync-run of module relaydescs of CollecTor. 2018-08-04 17:37:17,556 INFO o.t.m.c.s.SyncManager:57 Collecting Relay from collector2.torproject.org ... 2018-08-04 17:37:17,556 INFO o.t.d.i.DescriptorIndexCollector:46 Starting descriptor collection. 2018-08-04 17:37:17,556 INFO o.t.d.i.DescriptorIndexCollector:62 Indexing local directory /srv/collector.torproject.org/collector/sync/Relay-collector2.torproject.org. 2018-08-04 17:37:17,562 INFO o.t.d.i.DescriptorIndexCollector:73 Fetching remote index file https://collector2.torproject.org/index/index.json. 2018-08-04 17:37:18,053 INFO o.t.d.i.DescriptorIndexCollector:82 Fetching remote files from https://collector2.torproject.org. [...] 2018-08-05 18:27:13,331 INFO o.t.d.i.DescriptorIndexCollector:88 Deleting extraneous files from local directory sync/Relay-collector2.torproject.org. 2018-08-05 18:27:13,375 INFO o.t.d.i.DescriptorIndexCollector:93 Finished descriptor collection.
I'm not sure what to do here. The log doesn't reveal anything obviously going wrong. I didn't look at the code yet, so maybe there's an obvious way to prevent such issues in the future. Hmm.