Closed droyad closed 2 years ago
A customer running into aggressive mutexes on SSH targets: https://octopusdeploy.slack.com/archives/CNHBHV2BX/p1631304355207500
Leaving this for now. Let's wait for feedback on how our own dynamic workers perform with the recent changes to tool extraction mutexes and then consider the next steps.
There's a process wide mutex around package acquisition as well as a
FullIsolation
ScriptIsolation
mutex. The ScriptIsolation one exists for clustered scenarios. Having it means all other script's can't run blocking the acquisition as well as other deployments, particularly on workers.Could we make it named ClusterWideMutex only instead?
Or, likely even better, perhaps push it all the way down to the target/worker and use a named system mutex?
Historical Context: We do this so that we don't clash on the target machine with multiple deployments acquiring the same package. Pre-workers this blocking wasn't a big problem as simultaneous deployments were relatively rare.