[Gluster-users] Is it recommended for Glustereventsd be running on all nodes?

[Jeevan Patnaik]

Hi All,

I have implemented a webhook and attached to glustereventsd to listen to
events and to send alerts on critical events.

So, I categorized events manually critical, informational and warning.

We are interested in only events that can cause issue to end users like
BRICK_DISCONNECTED (reducing redundancy of the volume), QUORUM_LOST
(possible downtime of subvoume), QUOTA_CROSSES_SOFTLIMIT, AFR_SPLIT_BRAIN
etc. and not included any other events that are resulted while some one
does admin tasks like PEER_ATTACH.

And I see atleast some events are local to the node like PEER_ATTACH and
doesn't appear from other gluster node.

My idea is to run glustereventsd service only on a gluster admin node, to
avoid possible load on the storage serving nodes due to traffic caused by
webhook events.

So, my question is are there any events local to node, which will be missed
in admin node but are fatal to end users, assuming that the admin node will
always be running?

[Jeevan Patnaik]

Hi,

And Gluster version is 3.12.5.

Regards,
Jeevan.

Post by Jeevan Patnaik
Hi All,
I have implemented a webhook and attached to glustereventsd to listen to
events and to send alerts on critical events.
So, I categorized events manually critical, informational and warning.
We are interested in only events that can cause issue to end users like
BRICK_DISCONNECTED (reducing redundancy of the volume), QUORUM_LOST
(possible downtime of subvoume), QUOTA_CROSSES_SOFTLIMIT, AFR_SPLIT_BRAIN
etc. and not included any other events that are resulted while some one
does admin tasks like PEER_ATTACH.
And I see atleast some events are local to the node like PEER_ATTACH and
doesn't appear from other gluster node.
My idea is to run glustereventsd service only on a gluster admin node, to
avoid possible load on the storage serving nodes due to traffic caused by
webhook events.
So, my question is are there any events local to node, which will be
missed in admin node but are fatal to end users, assuming that the admin
node will always be running?

[Jeevan Patnaik]

Hi,

I get alerts from all nodes for PEER_DISCONNECT i.e., webhook config is
reflected on all nodes.

how to avoid this if at all we run glustereventsd on all nodes?

Post by Jeevan Patnaik
Hi,
And Gluster version is 3.12.5.
Regards,
Jeevan.

Post by Jeevan Patnaik
Hi All,
I have implemented a webhook and attached to glustereventsd to listen to
events and to send alerts on critical events.
So, I categorized events manually critical, informational and warning.
We are interested in only events that can cause issue to end users like
BRICK_DISCONNECTED (reducing redundancy of the volume), QUORUM_LOST
(possible downtime of subvoume), QUOTA_CROSSES_SOFTLIMIT, AFR_SPLIT_BRAIN
etc. and not included any other events that are resulted while some one
does admin tasks like PEER_ATTACH.
And I see atleast some events are local to the node like PEER_ATTACH and
doesn't appear from other gluster node.
My idea is to run glustereventsd service only on a gluster admin node, to
avoid possible load on the storage serving nodes due to traffic caused by
webhook events.
So, my question is are there any events local to node, which will be
missed in admin node but are fatal to end users, assuming that the admin
node will always be running?