model migration from 2.2.4 to 2.2.4 fails

Bug #1717860 reported by Paul Gear
14
This bug affects 2 people
Affects Status Importance Assigned to Milestone
Canonical Juju
Fix Released
High
Christian Muirhead

Bug Description

I tested model migration from a 2.2.2 to 2.2.4 controller today. It failed twice with the same status message: "migrating: aborted, removing model from target controller".

The controllers are both in the same tenant on an OpenStack cloud.

Revision history for this message
Paul Gear (paulgear) wrote :

Logs for the source and target controllers, along with a yaml status of the model which I attempted to migrate are at https://private-fileshare.canonical.com/~paulgear/lp1717860/

tags: added: canonical-is
Revision history for this message
Christian Muirhead (2-xtian) wrote :

Looking in the log reveals:
2017-09-18 06:39:40 ERROR juju.worker.migrationmaster.2780ee worker.go:279 model data transfer failed, model export failed: failed to read status history collection: Executor error during find command: OperationFailed: Sort operation used more than the maximum 33554432 bytes of RAM. Add an index, or specify a smaller limit.

The status history collection didn't have an index that was needed by migration - {"model-uuid", "-updated", "-_id"}
This was fixed in https://github.com/juju/juju/pull/7722 (so not in 2.2.2, but in 2.2.3+).

Revision history for this message
Christian Muirhead (2-xtian) wrote :

I think Paul's problem has been solved by upgrading to 2.2.4 - a workaround would have been to add the index in the mongo shell.

Changed in juju:
status: New → Fix Released
Revision history for this message
Paul Gear (paulgear) wrote :

I upgraded to 2.2.4 on the source controller and tried again. This failed also; logs uploaded to https://private-fileshare.canonical.com/~paulgear/lp1717860/1717860-source-try2.tar.gz

Revision history for this message
Paul Gear (paulgear) wrote :
summary: - model migration from 2.2.2 to 2.2.4 fails
+ model migration from 2.2.4 to 2.2.4 fails
Tim Penhey (thumper)
tags: added: new-york
Revision history for this message
Paul Gear (paulgear) wrote :

After the 3rd failed attempt at migration, the source controller became much less responsive - 260 seconds to run "juju models" on a controller with only 2 models: https://pastebin.canonical.com/199228/ Rebooting that controller got the time down to around 5 seconds.

Paul Gear (paulgear)
Changed in juju:
status: Fix Released → New
John A Meinel (jameinel)
Changed in juju:
status: New → Triaged
importance: Undecided → High
assignee: nobody → Christian Muirhead (2-xtian)
Revision history for this message
Paul Gear (paulgear) wrote :

Dug into this today with @2-xtian; new logs uploaded to https://private-fileshare.canonical.com/~paulgear/lp1717860/ - he found an issue with the OpenStack provider which resulted in the target controller not able to fully adopt the migrated model.

Revision history for this message
Christian Muirhead (2-xtian) wrote :

PR for 2.3-beta2, hopefully.
https://github.com/juju/juju/pull/7962

Changed in juju:
status: Triaged → Fix Committed
milestone: none → 2.3-beta2
Changed in juju:
status: Fix Committed → Fix Released
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.