On Wed, 9 Aug 2023, at 9:05 AM, Phil Dibowitz wrote: > On 8/7/23 07:24, Phil Dibowitz wrote: >> On 8/6/23 19:14, ellie timoney wrote: >>> Hmm. How much time elapses in the logs between the "updating recovery >>> stamp" and the first "assuming the worst"? I guess if something is >>> deleting the skipstamp file out from under Cyrus, that'll narrow down >>> when it's happening... >> >> assuming the worst shows up before that, and after. The first one in >> that log is before though:
Ah whoops, I meant the first "assuming the worst" to occur after the most recent "updating recovery stamp", not just the first one in the log. What I'm trying to see here is whether the skipstamp file is missing immediately after we think it's been created, or if it exists for a while but then disappears. You won't be able to see this easily with (z)grep... you'll probably want to open the log in something like less or vim (or whatever you're comfortable in), jump to the end, search backwards for "updating recovery stamp", make a note of the timestamp, then search forwards for "assuming the worst", and make a note of that timestamp too. Have you seen "DBERROR: write failed: filename=</path/to/db/skipstamp> ..."? If this error is happening, it'll be very soon after "updating recovery stamp", just once per restart. It means it's failing while trying to write the skipstamp file, which would explain why it doesn't exist, and the other details in the log line will tell us why. You haven't mentioned seeing this error, so I'm assuming this is not what's happening. Cheers, ellie ------------------------------------------ Cyrus: Info Permalink: https://cyrus.topicbox.com/groups/info/Tf92f39d4795cc515-M55014b411ac62540c0801aeb Delivery options: https://cyrus.topicbox.com/groups/info/subscription
