riak health check
Ali Rıza KELEŞ
ali.r.keles at gmail.com
Fri Oct 9 02:55:06 EDT 2015
On 8 October 2015 at 12:26, Sargun Dhillon <sargun at sargun.me> wrote:
> It's possible that during the strongly consistent joined, there was some
> leader instability. Do you have any logs of the event? Can you recreate the
> event? Also, the recommendation for SC, is to either turn off tree
> verification, or run with 7 nodes, and n=5.
Even I add two more nodes, in my case haproxy would get wrong state.
"GET /ping" is not valid method to identify whether the node is ready
to be responsible for cluster. Is that right?
What about parsing "GET /stats"? Has it any key to help me?
> On Thu, Oct 8, 2015 at 2:23 AM, Ali Rıza KELEŞ <ali.r.keles at gmail.com>
>> I have a riak cluster of 5 nodes, runinng strong consistent and behind
>> an HaProxy. Yesterday I needed to replace a node which was misplaced.
>> I removed the node and all was OK. Other nodes responded all requests
>> properly and my application continued to run well.
>> I created a new node and make it join the cluster. Application logged
>> some riak errors for 10 or 15 seconds, just after re-clustering. After
>> this period of errors, everything was OK again.
>> I have an idea why I encountered this stiuation but i am not sure.
>> HaProxy health checks got OK from the new node, unless it wasn't ready
>> or even it wasn't a member of riak cluster. And some requests
>> redirected by haproxy to this new alive node.
>> If this scenario is correct, then i need to make haproxy more clever
>> by tuning health check. "option httpchk GET /ping" is not enough to
>> identify a node is really ready for requests. Is there any fine tuned
>> way to achieve this.
>> Thanks for any suggestions.
>> Ali Rıza Keleş
>> riak-users mailing list
>> riak-users at lists.basho.com
Ali Rıza Keleş
More information about the riak-users