[Gluster-users] feature request - improved auto re-connect?

[Kingsley]

Hi,

(using CentOS 6.5 / Gluster 3.6.0beta3).

I've been testing various failure scenarios and have managed to upset
gluster by changing which bricks are viewable by the client.

I've created a volume gv0 with 2 replicas across 2 bricks
test1:/data/brick/gv0 and test2:/data/brick/gv0

I edited iptables on the brick servers test1 and test2 such that only
test2 was visible to client machine test5, and then made some changes.
Some hours later, I edited iptables on test1 and test2 so that now, only
test1 was visible to client machine test5. This seemed to upset it (see
transcript below).

Would it be possible for clients to cache a full list of brick servers
and then use that cached list to do $clever_things to maintain
connectivity / re-connect automatically?

Transcript follows (this was just after I switched iptables round on the
bricks after the test5 client had been using just the one accessible
brick for a while):

(I was sitting in a subdirectory of /mnt/gv0 initially):
test5# ls -l
ls: cannot open directory .: Transport endpoint is not connected

test5# df
df: `/mnt/gv0-slave': Transport endpoint is not connected
df: `/mnt/gv0': Transport endpoint is not connected
Filesystem           1K-blocks    Used Available Use% Mounted on
/dev/mapper/VolGroup-lv_root
                       5716804 1842760   3583640  34% /
tmpfs                   508140       0    508140   0% /dev/shm
/dev/xvda1              495844  121241    349003  26% /boot
test1:gv1              5716736 2299008   3127424  43% /mnt/gv1

test5# mount -t glusterfs test1:gv0 /mnt/gv0
ERROR: Mount point does not exist.
Usage:  mount.glusterfs <volumeserver>:<volumeid/volumeport> -o <options> <mountpoint>
Options:
man 8 mount.glusterfs

To display the version number of the mount helper:
mount.glusterfs --version

test5# cd /
test5# mount -t glusterfs test1:gv0 /mnt/gv0
ERROR: Mount point does not exist.
Usage:  mount.glusterfs <volumeserver>:<volumeid/volumeport> -o <options> <mountpoint>
Options:
man 8 mount.glusterfs

To display the version number of the mount helper:
mount.glusterfs --version

test5# ls -l /mnt/
ls: cannot access /mnt/gv0: Transport endpoint is not connected
ls: cannot access /mnt/gv0-slave: Transport endpoint is not connected
total 4
d????????? ? ?    ?       ?            ? gv0
d????????? ? ?    ?       ?            ? gv0-slave
drwxr-xr-x 3 root root 4096 Oct 24 12:29 gv1

test5# mount
/dev/mapper/VolGroup-lv_root on / type ext4 (rw)
proc on /proc type proc (rw)
sysfs on /sys type sysfs (rw)
devpts on /dev/pts type devpts (rw,gid=5,mode=620)
tmpfs on /dev/shm type tmpfs (rw)
/dev/xvda1 on /boot type ext4 (rw)
none on /proc/sys/fs/binfmt_misc type binfmt_misc (rw)
none on /proc/xen type xenfs (rw)
test4:gv0-slave on /mnt/gv0-slave type fuse.glusterfs (rw,default_permissions,allow_other,max_read=131072)
test1:gv0 on /mnt/gv0 type fuse.glusterfs (rw,default_permissions,allow_other,max_read=131072)
test1:gv1 on /mnt/gv1 type fuse.glusterfs (rw,default_permissions,allow_other,max_read=131072)

test5# umount /mnt/gv0-slave

test5# mount
/dev/mapper/VolGroup-lv_root on / type ext4 (rw)
proc on /proc type proc (rw)
sysfs on /sys type sysfs (rw)
devpts on /dev/pts type devpts (rw,gid=5,mode=620)
tmpfs on /dev/shm type tmpfs (rw)
/dev/xvda1 on /boot type ext4 (rw)
none on /proc/sys/fs/binfmt_misc type binfmt_misc (rw)
none on /proc/xen type xenfs (rw)
test1:gv0 on /mnt/gv0 type fuse.glusterfs (rw,default_permissions,allow_other,max_read=131072)
test1:gv1 on /mnt/gv1 type fuse.glusterfs (rw,default_permissions,allow_other,max_read=131072)

test5# umount /mnt/gv0
test5# ls -l /mnt
total 12
drwxr-xr-x 2 root root 4096 Sep 23 15:55 gv0
drwxr-xr-x 2 root root 4096 Oct 13 15:09 gv0-slave
drwxr-xr-x 3 root root 4096 Oct 24 12:29 gv1

-- 
Cheers,
Kingsley.