Fabian Arrotin wrote:

FM wrote:

...

Do you know whent Centos 5.3 will be ready ? Can I install my GFS2 FS with centos 5.2 and then "simply" upgrade to 5.3 without reinstallation ?

No, you'll have manual intervention because now in 5.3 gfs2 module is included by default in the 2.6.18-128.el5 kernel and isn't a external rpm to install.

I had a file system loss yesterday with gfs2, so have fun with that module.

I have an iscsi target which is reachable via two paths, and every time I took away one of the two paths, gfs2 panicked on me:

Mar 20 16:31:08 kavring iscsid: connection2:0 is operational after recovery (44 attempts) Mar 20 16:31:12 kavring multipathd: sdb: readsector0 checker reports path is down Mar 20 16:31:27 kavring last message repeated 3 times Mar 20 16:31:32 kavring kernel: GFS2: fsid=on3-ruby-app:htmldata.0: fatal: invalid metadata block Mar 20 16:31:32 kavring kernel: GFS2: fsid=on3-ruby-app:htmldata.0: bh = 132517 (magic number) Mar 20 16:31:32 kavring kernel: GFS2: fsid=on3-ruby-app:htmldata.0: function = gfs2_meta_indirect_buffer, file = fs/gfs2/meta_io.c, line = 334 Mar 20 16:31:32 kavring kernel: GFS2: fsid=on3-ruby-app:htmldata.0: about to withdraw this file system Mar 20 16:31:32 kavring kernel: GFS2: fsid=on3-ruby-app:htmldata.0: telling LM to withdraw Mar 20 16:31:32 kavring multipathd: sdb: readsector0 checker reports path is down Mar 20 16:31:33 kavring kernel: GFS2: fsid=on3-ruby-app:htmldata.0: withdrawn

After that you need to gfs2_fsck the system. With gfs it works as designed, I can reach the target via the other path and the file system stays online *and* uncorrupted.

If you get different results, I'd be interested to know.

Ralph