Do nodes in a cluster continuously ping a quorum server

[J.D._3]

I was under the impression that nodes in a SG cluster only communicate with the quorum server if a node reboots or needs to use it as a tiebreaker. Our configuration is as follows:

Nodes: RedHat Linux AS3,4,5
Quorum Server: RedHat Linux

ServicGuard: A.11.16.01-99
Quorum: qs-A.02.00.00-0

When doing a dump, I can see communication between one of the nodes and the quorum server. Please see below.

19:10:04.728297 IP abcla1a.57892 > qsla1.hacl-qs: P 50:60(10) ack 73 win 5840

19:10:04.768910 IP qsla1.hacl-qs > abcla1a.57892: . ack 60 win 46

19:10:04.769314 IP abcla1a.57892 > qsla1.hacl-qs: P 60:75(15) ack 73 win 5840

19:10:04.769330 IP qsla1.hacl-qs > abcla1a.57892: . ack 75 win 46

19:10:04.769413 IP qsla1.hacl-qs > abcla1a.57892: P 73:109(36) ack 75 win 46

19:10:04.769809 IP abcla1a.57892 > qsla1.hacl-qs: . ack 109 win 5840


Can someone clarify? Thanks in advanced.

[Steven E. Protter]

Shalom,

No SG nodes do not continuously ping the cluster.

They use a heart beat network, which is similar to ping to keep track of other nodes.

qsla1.hacl-qs

That is SG.

SEP

[J.D._3]

Hello Steven,

Thanks for the reply. The trace I have above is between one node (abcla1a) in the cluster and the quorum server (qsla1). I realize that there is a heartbeat between the nodes in the cluster that checks the health of each nodes in the cluster. How about the packets between nodes in the cluster and a quorum server? My thought is that nodes only checks with the quorum server to verify if a package should run in the primary or alternate node, to prevent split brain. When does a cluster node verify quorum with a quorum server?

[Matti_Kurkela · Accepted solution]

ServiceGuard is monitoring the presence of the quorum server, just to make sure it is available. If the quorum server becomes inaccessible when everything else is OK, the cluster nodes will output a warning to the system log.

The warning text is something like this: "The quorum server can not be accessed. If the heartbeat connections between the nodes fail now, the entire cluster may fail."

If there is a problem with accessing the quorum server, I think you'd appreciate getting a warning about the problem immediately.

The alternative would be seeing your entire cluster crash and then discovering that your quorum server has been unreachable because of a firewall configuration error for the last half a year or so...

MK

[J.D._3]

Matti,

Thanks for your reply. I finally looked into the config file after your post and did realize there is a QS_POLLING_INTERVAL setting and is described as:

The QS_POLLING_INTERVAL (microseconds) is the interval at which ServiceGuard checks to make sure the quorum server is running.

You're reply and this setting just validated my packet capture. Thanks.