Persisted shared ports during inplace update #9736
Conversation
scheduler/util.go
Outdated
| DiskMB: int64(update.TaskGroup.EphemeralDisk.SizeMB), | ||
| DiskMB: int64(update.TaskGroup.EphemeralDisk.SizeMB), | ||
| Ports: update.Alloc.AllocatedResources.Shared.Ports, | ||
| Networks: update.Alloc.AllocatedResources.Shared.Networks, |
There was a problem hiding this comment.
@nickethier pining if you happen to know, but noticed networks was dropped too, I'm guessing we want to carry over all fields
There was a problem hiding this comment.
I think so, but I think we'd want a copy of it. I wonder if we should just set Shared on line 730 to update.Alloc.AllocatedResources.Shared.Copy()?
There was a problem hiding this comment.
Yeah, its not clear to me why DiskMB is update.TaskGroup.EphemeralDisk.SizeMB instead of update.Alloc.AllocatedResources.Shared.DiskMB
There was a problem hiding this comment.
Ah I see, above Ports and Networks are set from the existing alloc because those cannot be updated inplace and is guarded by tasksUpdated.
The same is happening when setting the DiskMB here. So I think we can leave as is but would still add a .Copy() to the Networks.
AllocatedSharedResources were not being copied over to the new allocation struct the scheduler makes during inplace updates. This caused downstream issues after the plan was applied, namely the shared ports were dropped causing issues with service registration/deregistration. test that shared ports are preserved change log, also carry over shared network copy networks
drewbailey
force-pushed
the
b/persist-allocated-resources
branch
from
a481ae0
to
1cc88d7
Compare
AllocatedSharedResources were not being copied over to the new allocation struct the scheduler makes during inplace updates. This caused downstream issues after the plan was applied, namely the shared ports were dropped causing issues with service registration/deregistration. test that shared ports are preserved change log, also carry over shared network copy networks
|
I'm going to lock this pull request because it has been closed for 120 days ⏳. This helps our maintainers find and focus on the active contributions. |
Ensure that allocation AllocatedResources Shared Ports are not dropped during inplace updates.
When performing in place updates, the scheduler creates a shallow copy of the allocation and updates a few fields. It's not immediately clear to me why we do a shallow copy instead of something like
allocation.Copy(). The AllocatedSharedResources Ports were not copied over causing issues after the plan was applied.As a side effect, removing or updating a task group service stanza, resulting in an inplace update would error and prevent a service from being registered or deregistered during an inplace update. This issue was reported in #9360
This error was displayed in the logs like so:
fixes #9360
fixes #9735