Michael Eischer

@lego-2024 The `check successful on second attempt, original error pack` error along with lots of different data corruptions types is very likely caused by data corruption in the CPU or...

> After running the Prime95 (version [30.19](https://download.mersenne.ca/gimps/v30/30.19/p95v3019b20.linux64.tar.gz)) for 24 hours, the system has completed the torture test without errors/warnings, suggesting that it is reliable and functioning as expected. And yet...

> The distinction between the original run and the second and third executions is that the message "check successful on second attempt" was not present in the latter two Hmm,...

The results are somewhat weird: the `pack [...] contains 1 error` are largely consistent with those from the initial list of errors. This means that they definitely got corrupted during...

> It looks like the main issue was caused by me. Upon examining the repository folder, I discovered another repository with the same name as the original. I don't really...

The behavior in the logs is the very definition of bitflips. I unfortunately don't have a good idea how to properly diagnose this any further.

Somewhat related to #4122 . That option might also be interesting for other commands that deal with multiple snapshots.

> 1. The copy uses _$RESTIC_HOST_ as default, ie no way to select all hosts (wildcard) Not sure what you mean. Just don't set the hostname if you want to...

> Are there any plans to continue development of rest-server as an independent backend? There's no plan to abandon rest-server. But yes development has slowed to a crawl as I'm...

The option is used to create a CPU usage profile for rest-server. See https://go.dev/blog/pprof for more context of the profiling built into the go runtime. You can use it using...