ERROR log from space-importer after mgopurge

Bug #1700731 reported by Peter Sabaini on 2017-06-27
8
This bug affects 1 person
Affects Status Importance Assigned to Milestone
juju
High
Unassigned

Bug Description

After running a mgopurge (all stages) with state server shutdown and subsequent start I'm now getting this log messages repeated over and over (for the last ~16h or so):

2017-06-27 09:25:47 ERROR juju.worker.dependency engine.go:547 "space-importer" manifold worker returned unexpected error: adding subnet failed: adding subnet "1.2.3.4/5": subnet "1.2.3.4/5" already exists

Subnet is redacted, but it is an existing and valid subnet, albeit not in use currently.

Version info: juju 2.1.3, provider is maas 2.2, mgopurge 1.5

I can see the folder holding the log file, but I'm unable to download the
log file itself.

Can you tell us what cloud this is? Offhand I would expect this to be an
issue of restarting the controller, not because of running mgopurge.

The spaces discovery worker in 2.1.3 runs at startup (and clearly is
bouncing due to being unhappy about this particular issue). In 2.2 it only
runs on initial bootstrap or add-model or when triggered by the user (juju
reload-spaces).

One question is if a given subnet CIDR has been removed and then added
again. Which might cause us to error with a conflict.

So the particular 'worker is erroring over and over in the log file' should
be fixed in 2.2, but the actual "handle when subnets are removed and
re-added" is probably not.

On Tue, Jun 27, 2017 at 8:30 PM, Peter Sabaini <email address hidden>
wrote:

> I've uploaded machine-0 log here: https://private-
> fileshare.canonical.com/~sabaini/Lo4ahl2xohk3quoo/lp1700731/
>
> --
> You received this bug notification because you are subscribed to juju.
> Matching subscriptions: juju bugs
> https://bugs.launchpad.net/bugs/1700731
>
> Title:
> ERROR log from space-importer after mgopurge
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/juju/+bug/1700731/+subscriptions
>

Peter Sabaini (peter-sabaini) wrote :

John,

apologies for the perm on the logfile, they're now updated.

This is Mitaka cloud on xenial; both juju and maas are running in a KVM each. Did you need any specific details?

I think indeed during initial setup this subnet had been deleted and re-added - so that would be an explanation

John A Meinel (jameinel) wrote :

At this point we don't support having a subnet go away and come back with a different identifier. It would be possible to edit the database directly, and we have some plans around improving our space support (being able to add/remove/update existing definitions), but that work has not been started yet. It may be that we can get to some of this in a 2.3 time frame (in the next month or two).

To handle the immediate issue, we either need to edit the database, or rebootstrap.

Changed in juju:
importance: Undecided → High
status: New → Triaged
To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Other bug subscribers