Some gluster questions

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hello,
we are playing arround gluster and some technical and theoretical questions came up:

1. Is it wise to let gluster do the local striping over disks/bricks?

We wanted to use ZFS (as for its simplified management and nice features) but as there are issues installing OpenSolaris on HP DL380 (OS doesnt support install from USB (through ILO)) we ended up making a bunch of raid1 (mirror) volumes and mounting them together with Gluster unify. But I am not sure which way would it be better - to make a large raid 6/10 volume (I personally dont like hardware raids which involve more than 6 disks) or bind the mirrors together with LVM and then export through Gluster or use it this way?


2. What is the correct way of adding a new local brick to (allready existing/filled) unify storage?

I have defined a new brick volume and added it to unify subvolumes restarted both the client and server but the ALU scheduler still doesnt create any new files on the fresh brick


The unify config:

volume c1d1
 type storage/posix
 option directory /data/raw_c1d1
end-volume

....
....

volume unify
 type cluster/unify
subvolumes c1d1 c1d2 c1d3 c1d4 c1d5 c1d6 c1d7 c1d8 c2d0 c2d1 c2d2 c2d3 c2d4 c2d5 c2d6
 option namespace ns
 option scheduler alu
 option alu.limits.min-free-disk  5%
 option alu.limits.max-open-files 10000
option alu.order disk-usage:read-usage:write-usage:open-files-usage:disk-speed-usage
 option alu.disk-usage.entry-threshold 10GB
 option alu.disk-usage.exit-threshold  60MB
 option alu.open-files-usage.entry-threshold 1024
 option alu.open-files-usage.exit-threshold 32
 option alu.stat-refresh.interval 10sec
end-volume


Could it be that the alu.disk-usage.entry-threshold is too high? And the new volume doesnt get used until there is a 10Gb difference to others? (probably have to play arround first rather than asking questions)



3. Is there another way to mount the volume locally not using TCP?

We are transfering files from an older disk storage and setting up fuse kernel module and gluster client there is problematic. So we have mounted the exported gluster volume just back on the new storage server and transfering the files to the mountpoint (tar/ssh).

Client config on the new box looks only like:

volume remote
 type protocol/client
 option transport-type tcp/client
 option remote-host 127.0.0.1
 option remote-subvolume storage
end-volume


As I have seen people and in docs saying that you can load the translators anywhere - is it possible just to copy over all the volumes to client config (without the "protocol/server" volume) and "mount" the unify partition without the 'glusterfsd' at all?


In this case scenario (mounting locally over TCP) should be the performance translators like io-threads / write-behind loaded also on client or is it enough to have them only on server (as they are the same box anyway)? Should the translators be on both client/server when there are different boxes?


wbr
Reinis Rozitis





[Index of Archives]     [Gluster Users]     [Ceph Users]     [Linux ARM Kernel]     [Linux ARM]     [Linux Omap]     [Fedora ARM]     [IETF Annouce]     [Security]     [Bugtraq]     [Linux]     [Linux OMAP]     [Linux MIPS]     [eCos]     [Asterisk Internet PBX]     [Linux API]

  Powered by Linux