[Gluster-users] CHANGELOGs and new geo-replica sync taking forever

[Brian Ericson]

Thanks for the reply!

On 11/04/2015 06:46 PM, Wade Fitzpatrick wrote:
> I also had problems getting geo-replication working correctly and
> eventually gave it up due to project time constraints.
>
> What version of gluster?

I'm using 3.7.5 on all nodes.

> What is the topology of x, xx, and xxx/xxy/xxz?

x & xx are both "stripe 2 replica 2" (4 hosts each); xxx, xxy, and xxz 
are all "replica 2" (2 hosts each).  16 machines total.

>
> I tried a 2x2 stripe-replica with geo-replication to a 2x1 stripe using
> 3.7.4. Starting replication with 32 GB of small files never completed,
> it failed several times. Starting replication with an empty volume then
> filling it with a rate limit of 2000k/s managed to keep sync until
> completion but could not handle the rate of change under normal usage.

I'm not expecting a lot of change -- new files are added infrequently 
and all files are written once and never modified.  I'm ...perplexed... 
by the sheer number of CHANGELOG files and see "getting over the hump" 
to be problematic.

In fact, for every 3 CHANGELOG files that get transferred (xx -> xxz), 
I'm seeing 2 more get created on xx; transferring 3 files is only 
netting me one!!!

>
> On 5/11/2015 3:30 AM, Brian Ericson wrote:
>> tl;dr -- geo-replication of ~200,000 CHANGELOG files is killing me...
>> Help!