Recovery backend always gives timeout even though cluster is healthy
Hi, I was trying to test out the new recovery backend, rados_kv. I am using nfs-ganesha v2.7-dev13 with latest ceph master. I am testing using vstart cluster. But everytime I get timeout "Failed to connect to cluster: -110". The cluster is healthy. I can also give commands using the rados cli. Any pointers on what could be failing here? Also, while working on this, I realized the default timeout (client_mount_timeout) is 300 in seconds, and there are 10 tries. So, in worst case timeout occurs after 50 mins. The option, "client_mount_timeout" can be set in ceph.conf, but it seems like its also possible to set in the ganesha config and pass it in function rados_conf_set(). [1] Will it be useful to have such option in ganesha.conf? [1] https://tracker.ceph.com/issues/6507 ------ Supriti Singh SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nürnberg)
On Mon, 2018-05-28 at 16:30 +0200, Supriti Singh wrote:
Hi,
I was trying to test out the new recovery backend, rados_kv. I am using nfs-ganesha v2.7-dev13 with latest ceph master. I am testing using vstart cluster. But everytime I get timeout "Failed to connect to cluster: -110". The cluster is healthy. I can also give commands using the rados cli. Any pointers on what could be failing here?
Also, while working on this, I realized the default timeout (client_mount_timeout) is 300 in seconds, and there are 10 tries. So, in worst case timeout occurs after 50 mins. The option, "client_mount_timeout" can be set in ceph.conf, but it seems like its also possible to set in the ganesha config and pass it in function rados_conf_set(). [1] Will it be useful to have such option in ganesha.conf?
It sounds like you have a communications issue, or perhaps your ganesha isn't using the correct ceph.conf. I'd sniff traffic to see what it's doing, or look at lsof output to see what sockets it's trying to open. -- Jeff Layton <jlayton@poochiereds.net>
Jeff Layton <jlayton@poochiereds.net> 05/28/18 5:47 PM >>> On Mon, 2018-05-28 at 16:30 +0200, Supriti Singh wrote: Hi,
I was trying to test out the new recovery backend, rados_kv. I am using nfs-ganesha v2.7-dev13 with latest ceph master. I am testing using vstart cluster. But everytime I get timeout "Failed to connect to cluster: -110". The cluster is healthy. I can also give commands using the rados cli. Any pointers on what could be failing here?
Also, while working on this, I realized the default timeout (client_mount_timeout) is 300 in seconds, and there are 10
Its using the right ceph.conf. I attached ganesha with gdb, and its always looping around https://github.com/ceph/ceph/blob/master/src/mon/MonClient.cc#L499 , and then it times out. I don't know the monitor code well enough to understand what is it trying to do here. I am running vstart with option -X, so it should not be looking for cephx credentials. ------ Supriti Singh SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nürnberg) tries. So, in worst case timeout occurs after 50 mins. The option, "client_mount_timeout" can be set in ceph.conf, but it seems like its also possible to set in the ganesha config and pass it in function rados_conf_set(). [1] Will it be useful to have such option in ganesha.conf?
It sounds like you have a communications issue, or perhaps your ganesha isn't using the correct ceph.conf. I'd sniff traffic to see what it's doing, or look at lsof output to see what sockets it's trying to open. -- Jeff Layton <jlayton@poochiereds.net> _______________________________________________ Devel mailing list -- devel@lists.nfs-ganesha.org To unsubscribe send an email to devel-leave@lists.nfs-ganesha.org
participants (2)
-
Jeff Layton -
Supriti Singh