WIP: hydra-queue-runner: do not assume reservation
is available when build step throws unexpectedly #1
Loading…
Reference in a new issue
No description provided.
Delete branch "raito/hydra:rsv-nullpo"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
reservation
might be null when an uncaught exception has been thrown.We copy the data we need to report the error.
Not sure if this is the right way to go about this fix.
reservation
is available when build step throws unexpectedlyI don't think it passes all the tests, one is failing (flakiness, idk?).
And it's the LDAP test…
Yeah that looks correct - a while back I made it so machines drop their reservations early to not hog build slots while copying stuff around, and I'm guessing we've been encountering exceptions that happen on the "copying outputs" step?
(See
5c3e508e55
)@ -40,2 +41,4 @@
res = doBuildStep(destStore, reservation, activeStep);
} catch (std::exception & e) {
if (!reservation) {
printMsg(lvlError, "machine '%s' has been released unexpectedly", machineSshName);
Not unexpectedly. See my other comments.
Yep, makes sense. Will change it, to a
lvlTalkative
maybe?a8a5dea411
to1f9d8dc0a6
The LDAP test still fails deterministcally, I have no idea why.
Fixed in
fb9e29d4
.Pull request closed