Conversation
* Use /ram/imgsrv-metrics for a consistent location for shared location for updating metrics (/tmp in Debian 13 is too transient and has caused issues) * Remove Utils::MonitorRun -- this was pretty hacky and we haven't ended up using the metrics from it. If this continues to cause issues, we may want to back off on metrics here entirely and instead analyze metrics from the logs until we have a better sense of what to record & a stable way to record it.
|
I'll want to verify this works as expected in dev before merging it. I'm not sure what the interaction will be between the persistent PSGI imgsrv and other things that might try to update the stats, and I don't remember how exactly this was working before. If we can't get it working reliably, we can just turn it off. |
|
Probably not super urgent as @rrotter has put another fix in place to avoid the Debian 13 behavior of trying to automatically clean out old stuff in |
|
|
||
| my @yes; | ||
| # Pipe echo to unzip so it won't hang on a user prompt when ramdisk is full | ||
| push @yes, "echo", "n"; |
There was a problem hiding this comment.
Is "yes" saying "no" to unzip? I haven't ventured much into this corner of imgsrv, so it raises the question for me, what question is unzip prompting us to answer? And what is the resulting behavior when we say "n" to whatever question this is? (I realize the handling of the unzip call is peripheral to what we're really doing here, but since we're here...)
There was a problem hiding this comment.
That's a good question...
moseshll
left a comment
There was a problem hiding this comment.
See comment-in-passing. Nothing here fills me with dread. APPROVE
|
Continuing to use I'd like |
Since imgsrv is already using |
Partially reverts #13
Use /ram/imgsrv-metrics for a consistent location for shared location for updating metrics (/tmp in Debian 13 is too transient and has caused issues)
Remove Utils::MonitorRun -- this was pretty hacky and we haven't ended up using the metrics from it.
If this continues to cause issues, we may want to back off on metrics here entirely and instead analyze metrics from the logs until we have a better sense of what to record & a stable way to record it.