Mark Mielke wrote: > On 10/21/2009 06:42 PM, Tomasz Chmielewski wrote: >> Tomasz Chmielewski wrote: >>> How can I make the glusterfs client re-read its configuration file? >>> >>> I added some more glusterfs servers and would like the clients to use >>> them. >>> >>> How can I do it, without unmounting and mounting everything on >>> clients from scratch? >> >> Nobody knows? >> >> Should I assume it's impossible and killing a gluster fuse client and >> starting it again is the only way to "see" new servers? > > How do you make an NFS client, or AutoFS see new clients? With NFS, it's pretty easy - resize the filesystem on the server, it will be automatically seen by the client. For example: client# df -h | grep src 192.168.8.64:src 2,0G 1,1G 814M 58% /mnt/1 server# lvextend -L+500M /dev/san5/src Extending logical volume src to 2.49 GB Logical volume src successfully resized server# resize2fs /dev/san5/src resize2fs 1.40-WIP (14-Nov-2006) Filesystem at /dev/san5/src is mounted on /mnt/src; on-line resizing required old desc_blocks = 1, new_desc_blocks = 1 Performing an on-line resize of /dev/san5/src to 652288 (4k) blocks. The filesystem on /dev/san5/src is now 652288 blocks long. client# df -h | grep src 192.168.111.177:src 2,5G 1,1G 1,3G 47% /mnt/1 No way to do a similar thing with glusterfs (by adding new servers to the common mount point on the client)? That does not scale as well as advertised... ;) > Unmount and mount is the only thing that makes sense? For that, you have to stop everything which uses that mount point on clients. Which means down time and is usually not desired. > Changing a live file mount seems pretty dangerous... You can do a live resize of i.e. ext3/4 filesystems with resize2fs (as I shown above). Nothing really dangerous here, provided your machine doesn't crash in the meantime. And with a networked filesystem resize, I don't see this problem at all (and with NFS, it just works). -- Tomasz Chmielewski http://wpkg.org
