Hi Mariusz,
You haven't mention pcs version you are running. Based on you mentioning
running Pacemaker 2, I suppose you are running pcs 0.10.x. The text
bellow applies to pcs 0.10.x.
Pcs doesn't depend on or use corosync.conf when connecting to other
nodes. The reason is pcs must be able to connect to nodes not specified
in corosync.conf, e.g. when there is no cluster created yet.
Instead, pcs has its own config file mapping node names to addresses.
The easiest way to set it is to specify an address for each node in 'pcs
host auth' command like this:
pcs host auth <node1> addr=<ip1> <node2> addr=<ip2> ...
Specifying addresses is not mandatory. If the addresses are omitted, pcs
uses node names as addresses. See man pcs for more details.
To fix your issues, run 'pcs host auth' and specify all nodes and their
addresses. Running the command on one node of your cluster should be enough.
Regards,
Tomas
Dne 10. 08. 20 v 14:21 Mariusz Gronczewski napsal(a):
Hi,
Pacemaker 2, the current setup is
* management network with host's hostname resolving to host's
management IP
* cluster network for Pacemaker/Corosync communication
* corosync set up with node name and IP of the cluster network
pcs status shows both nodes online, added config syncs to the other
node etc. but pcs cluster status shows one node being offline.
After a look in firewall logs it appears all of the communication is
going just fine on the cluster network but PCS tries to talk with PCSD
on port 2224 via *management* network instead of using IP set as
ring0_addr in corosync
Is "just use host's hostname regardless of config" something normal ?
Is there a separate setting to pcs about which IP it should use ?
Regards
_______________________________________________
Manage your subscription:
https://lists.clusterlabs.org/mailman/listinfo/users
ClusterLabs home: https://www.clusterlabs.org/
