restarted node tells me about to handoff partitions (changing format to v1)

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

restarted node tells me about to handoff partitions (changing format to v1)

Effenberg, Simon
Hi,

after an upgrade from 1.3.1 to 1.4.2 I changed the object format from
v0 to v1 and began a rolling restart. But restarting the first node
shows me some weird outputs:

'riak@10.46.109.201' waiting to handoff 132 partitions

this is the "restarted" node and the partitions is switching between
121 and 132 .. after some time it decreases to something between 110
and 120.. and after some more time it is now decreasing slowly.

My question is: why/how can it be that the restarted node has stuff to
handover? Shouldn't it be the other way around (there were no handovers
in the "riak-admin transfers" before the restart).

My guess would be something with AAE which has had something in the
queue which didn't show up in the normal operation? Or do I miss
something?

Cheers
Simon

_______________________________________________
riak-users mailing list
[hidden email]
http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com
Reply | Threaded
Open this post in threaded view
|

Re: restarted node tells me about to handoff partitions (changing format to v1)

Shane McEwan-2
On 16/12/13 09:56, Simon Effenberg wrote:

> after an upgrade from 1.3.1 to 1.4.2 I changed the object format from
> v0 to v1 and began a rolling restart. But restarting the first node
> shows me some weird outputs:
>
> 'riak@10.46.109.201' waiting to handoff 132 partitions
>
> this is the "restarted" node and the partitions is switching between
> 121 and 132 .. after some time it decreases to something between 110
> and 120.. and after some more time it is now decreasing slowly.
>
> My question is: why/how can it be that the restarted node has stuff to
> handover? Shouldn't it be the other way around (there were no handovers
> in the "riak-admin transfers" before the restart).

I've seen this behaviour before, even in earlier versions. I assume that
when a Riak node starts up it needs to check each of its partitions
against the copies on the other nodes in order to determine if it needs
to send or receive updates. So the message should really say something
like: "waiting to check 132 partions."

The corresponding messages in the log are slightly clearer:

hinted_handoff transfer of riak_kv_vnode from 'riak@10.2.1.1'
485264215520026359559817229612828346370133852160 to 'riak@10.2.1.2'
485264215520026359559817229612828346370133852160 completed: sent 0
objects in 0.04 seconds

The fact that it sent 0 objects tells me that it didn't actually need to
repair anything.

Shane.


_______________________________________________
riak-users mailing list
[hidden email]
http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com