Bug 1328636

+++ This bug was initially created as a clone of Bug #1260630 +++

Technical Guide in 3.5.z, section 5.9 Migration:

A bandwidth limit of 30 Mbps is imposed on each virtual machine migration. A migration will timeout after a certain amount of time has passed. The timeout happens after either 300 seconds, or after the amount of virtual machine memory in Mb divided by 2048 Mb multiplied by 300 seconds, whichever is larger.
By default, concurrent outgoing migrations are limited to one per CPU core per host, or 5, whichever is smaller


^^ defaults are very outdated, ever since IIRC 3.2 we have a limit of 32MiBps (megabytes/s), concurrent migration has a limit of 3, there is a "progress" timeout of 150s (stalling, when migration doesn't progress it's aborted), 'migration_max_time_per_gib_mem' is 64s

--- Additional comment from Sandro Bonazzola on 2015-10-26 22:36:30 EST ---

this is an automated message. oVirt 3.6.0 RC3 has been released and GA is targeted to next week, Nov 4th 2015.
Please review this bug and if not a blocker, please postpone to a later release.
All bugs not postponed on GA release will be automatically re-targeted to

- 3.6.1 if severity >= high
- 4.0 if severity < high

--- Additional comment from Andrew Dahms on 2015-11-18 22:55:38 EST ---

Clearing needinfo request raised against the default assignee.

--- Additional comment from Yaniv Dary on 2016-03-16 22:55:19 EST ---

Is this still correct for 3.6?

--- Additional comment from Michal Skrivanek on 2016-03-16 23:03:29 EST ---

yes. until bug 1252426 gets in

--- Additional comment from Lucy Bopf on 2016-03-22 16:13:34 EST ---

Assigning to Tahlia for review.

--- Additional comment from Tahlia Richardson on 2016-04-14 09:53:57 EST ---

> "progress" timeout of 150s (stalling, when migration doesn't progress it's
> aborted), 'migration_max_time_per_gib_mem' is 64s

Just to clarify this last point, does this mean that "The timeout happens after either 300 seconds, or after the amount of virtual machine memory in Mb divided by 2048 Mb multiplied by 300 seconds, whichever is larger." should now be "The timeout happens after either 150 seconds, or after 64 seconds per Gb of virtual machine memory, whichever is larger."?



The paragraph would then read in full: "A bandwidth limit of 32 Mbps is imposed on each virtual machine migration. A migration will timeout after a certain amount of time has passed. The timeout happens after either 150 seconds, or after 64 seconds per Gb of virtual machine memory, whichever is larger.
By default, concurrent outgoing migrations are limited to one per CPU core per host, or 3, whichever is smaller." Is this correct?

--- Additional comment from Michal Skrivanek on 2016-04-14 17:18:46 EST ---

(In reply to Tahlia Richardson from comment #6)
> > "progress" timeout of 150s (stalling, when migration doesn't progress it's
> > aborted), 'migration_max_time_per_gib_mem' is 64s
> 
> Just to clarify this last point, does this mean that "The timeout happens
> after either 300 seconds, or after the amount of virtual machine memory in
> Mb divided by 2048 Mb multiplied by 300 seconds, whichever is larger."
> should now be "The timeout happens after either 150 seconds, or after 64
> seconds per Gb of virtual machine memory, whichever is larger."?
> 
> 
> 
> The paragraph would then read in full: "A bandwidth limit of 32 Mbps is
> imposed on each virtual machine migration. A migration will timeout after a
> certain amount of time has passed. The timeout happens after either 150
> seconds, or after 64 seconds per Gb of virtual machine memory, whichever is
> larger.
> By default, concurrent outgoing migrations are limited to one per CPU core
> per host, or 3, whichever is smaller." Is this correct?

It's 32MiBps (megabytes/s), not Mbps. Please make the distinction clear.
150s timeout describes only a situation when migration is not progressing. It's not a total limit, it's a "stalling" time limit, counting time when the amount of required memory to transfer increases or doesn't change as opposed to normal situation when it progresses from 0% to 100%. That 64s per 1GB is correct and it is indeed total/absolute limit.

Also, please note a bug 1323952 where we changed some of the values in 3.6.6 (bandwidth is 52MiBps, concurrent ongoing migrations are limited to 2, and stalling timeout is 240s). It's always good to refer e.g. https://access.redhat.com/solutions/744423 for more details about tuning migration for more details about the various settings and values

Thanks

--- Additional comment from Tahlia Richardson on 2016-04-20 10:00:34 EST ---

Thanks Michal. I'll add these updates for the next 3.5 and 3.6 updates, and clone a new bug for the 3.6.6 changes so they don't go out too early.