Hi Sean - yes we are using HAProxy - and there are health checks.
[image: photo]
*Jason Ryan*
VP Engineering
Trustev
Real Time, Online Identity Verification
email: jason.r...@trustev.com
skype: jason_j_ryan
web: www.trustev.com
Trustev Ltd, 2100 Cork Airport Business Park, Cork, Ireland.
On 4
We're seeing regular errors around indexes trying to be created - but they
already exist - indexes are only ever created manually, and no new indexes
have been created in quite a while. We have however added new nodes and
removed other nodes due to some node failures.
We also have large object war
Are there any other significant error messages in the logs? I find it hard
to believe that that specific crash would make your cluster unusable.
On Wed, Mar 4, 2015 at 10:07 AM, Jason Ryan wrote:
> This is what we have on our haproxy right now:
>
> backend riak_rest_backend
> mode
This is what we have on our haproxy right now:
backend riak_rest_backend
mode http
option httplog
balanceroundrobin
option httpchk GET /ping
server EU-RIAK-5 10.0.4.12:8098 weight 1 maxconn 1024 check
We actually use /ping already for the health check
[image: photo]
*Jason Ryan*
VP Engineering
Trustev
Real Time, Online Identity Verification
email: jason.r...@trustev.com
skype: jason_j_ryan
web: www.trustev.com
Trustev Ltd, 2100 Cork Airport Business Park, Cork, Ireland.
On 4 March 2015 at 1
We recently saw this on the PB interface as well and have mitigated it [1],
but we have not addressed it on HTTP. In the meantime, I would recommend
changing your health check to use a regular request, for example:
option httpchk /ping
[1] https://github.com/basho/riak_api/pull/54
On Wed, Mar 4,
Are you using a load balancer in front of Riak, like HAProxy, with health
checks turned on?
On Wed, Mar 4, 2015 at 7:24 AM, Jason Ryan wrote:
> Hi all,
>
> We're getting the following crash report messages in the log very
> regularly in the last day - it renders our Riak Cluster pretty much
> un
