| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Problem:
In unit file of TA process we have been using ta-vol as
volume id and also ta-vol-server.transport.socket.listen-port=24007
In volume file for TA process we only consider volname
as "ta" and not as "ta-vol". That's why it was not able to assign
this port number to ta process as in volume file it will try to
find server xlator as ta-vol
volume ta-server <<<<<<<<< not ta-vol-server
46 type protocol/server
47 option transport.listen-backlog 10
48 option transport.socket.keepalive-count 9
49 option transport.socket.keepalive-interval 2
50 option transport.socket.keepalive-time 20
51 option transport.tcp-user-timeout 0
52 option transport.socket.keepalive 1
53 option auth.addr./mnt/thin-arbiter.allow *
54 option auth-path /mnt/thin-arbiter
55 option transport.address-family inet
56 option transport-type tcp
57 subvolumes ta-io-stats
58 end-volume
Solution:
Provide "ta" as vol id for the command which Unit file
is going to execute.
Also, made changes in setup-thin-arbiter.sh to correctly
identify the directory of Unit file irrespective of the location from
where we are executing this script.
Change-Id: Ia7bbccdc0304e7dfaaa732bebb726fba731d1d33
fixes: bz#1716766
Signed-off-by: Ashish Pandey <aspandey@redhat.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Discussion on thin arbiter volume -
https://github.com/gluster/glusterfs/issues/352#issuecomment-350981148
Main idea of having this rpm package is to deploy thin-arbiter
without glusterd and other commands on a node, and all we need
on that tie-breaker node is to run a single glusterfs command.
Also note that, no other glusterfs installation needs
thin-arbiter.so.
Make sure RPM contains sample vol file, which can work by default,
and a script to configure that volfile, along with translator image.
Change-Id: Ibace758373d8a991b6a19b2ecc60c93b2f8fc489
updates: bz#1674389
Signed-off-by: Amar Tumballi <amarts@redhat.com>
Signed-off-by: Ashish Pandey <aspandey@redhat.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
If this option is "ON' and if there is a lock contention for
a file from 2 clients/self-heal daemon, an up call will
be sent to all the clients holding that locks. Clients
can handle this notification and may release the locks
at the right time asap.
If this option is ON, we get "instant" lock contention
notification on client side.
In next version of thin arbiter, we need to implement
lock contention notification handling and two domain
locking to synchronize update/access of xattrs of thin arbiter
replica id file. This option is an efficient way to find
out lock contention. In this case it is also helping to
implement two domain locking for replica id file.
Change-Id: I69883898d52feb425dfceb31cf12fa2157aaa314
fixes bz#1609207
Signed-off-by: Ashish Pandey <aspandey@redhat.com>
|
|
|
1 - setup-thin-arbiter.sh
To setup directory path for thin-arbiter. This is the path where replica id
file will be placed and used by TA process.
By default it will be /mnt/thin-arbiter. It is better to use default path
and once set it should not be changed.
2 - gluster-ta-volume.service
This is the Unit file used by systemd to start and keep the thin-arbiter
process running.
3 - thin-arbiter.vol
This is the default volume file for thin-arbiter. It will be used to start
thin-arbiter process. This should not be modified by user except the
directory path which can be modified using [1].
Change-Id: I68f2164f24adfa326c1d1a7a98c6ac4ff04ed553
fixes: bz#1579786
Signed-off-by: Ashish Pandey <aspandey@redhat.com>
|
