Tomasz Chmielewski
2011-May-27 22:35 UTC
[Gluster-users] not able to use gluster if one of the nodes offline
Hi, I'm using glusterfs 3.1.4 and just discovered that the whole is unusable if the second gluster server dies (using distributed replica). I have two gluster servers, called web3-int and web4-int. Whenever web4-int is not reachable, I'm not able to start glusterfs on web3-int. On web3-int, it fails with the following info in the log: glusterd: cannot resolve brick: web3-int:/home/gluster-data The workaround for this problem was "easy" (well, after spending some time debugging...) - manually adding a file with web3-int's UID to /etc/glusterd/peers directory. Why manually? Because according to "gluster peer probe" command, doing this for localhost is not needed: # gluster peer probe web3-int Probe on localhost not needed However, without localhost's file in /etc/glusterd/peers directory, glusterd on web3-int is not able to work properly, if web4-int is down. Is it expected? -- Tomasz Chmielewski http://wpkg.org
Mohit Anchlia
2011-May-27 23:21 UTC
[Gluster-users] not able to use gluster if one of the nodes offline
There seems to be some problem with your setup. Can you look at your /etc/hosts file and see if it's setup correctly on web3-int? Also, post output of gluster peer status, gluster volume info, ping web3-int, ping web4-int from the surviving node. On Fri, May 27, 2011 at 3:35 PM, Tomasz Chmielewski <mangoo at wpkg.org> wrote:> Hi, > > I'm using glusterfs 3.1.4 and just discovered that the whole is unusable if > the second gluster server dies (using distributed replica). > > I have two gluster servers, called web3-int and web4-int. > > Whenever web4-int is not reachable, I'm not able to start glusterfs on > web3-int. > > On web3-int, it fails with the following info in the log: > > glusterd: cannot resolve brick: web3-int:/home/gluster-data > > > > The workaround for this problem was "easy" (well, after spending some time > debugging...) - manually adding a file with web3-int's UID to > /etc/glusterd/peers directory. Why manually? Because according to "gluster > peer probe" command, doing this for localhost is not needed: > > # gluster peer probe web3-int > Probe on localhost not needed > > > However, without localhost's file in /etc/glusterd/peers directory, glusterd > on web3-int is not able to work properly, if web4-int is down. > > > Is it expected? > > > -- > Tomasz Chmielewski > http://wpkg.org > _______________________________________________ > Gluster-users mailing list > Gluster-users at gluster.org > http://gluster.org/cgi-bin/mailman/listinfo/gluster-users >