gluster-devel
[Top][All Lists]
Advanced

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

Re: [Gluster-devel] 2.0.0 Finally Happened?


From: ender
Subject: Re: [Gluster-devel] 2.0.0 Finally Happened?
Date: Mon, 04 May 2009 15:01:30 -0700
User-agent: Thunderbird 2.0.0.21 (X11/20090409)

Gordan Bobic wrote:
Thanks for pointing it out. Since these issues were mentioned:

1) Self-heal of a file that does not exist on the first subvolume

Is this affected by read-subvolume and favourite-child options? Does

It effects me and i only use "subvolumes". Using read-subvolume and or favourite child 
most likely only changes which node is the "first" node. Meaning the bug does not go 
away, its just server 'XYZ' can never go offline instead of server 'ABC'. I am just guessing here.


this lead to the file being clobbered on "ls -la", since the directory metadata is going to be inconsistent (at least different modification time since a file was created on one of the secondary nodes)? Is there a way to heal this file without knowing it's name (and knowing it's name isn't always feasible if it becomes invisible once the primary server comes back up!)? If it shows up at least on the server where it exists, then doing a periodic "ls -laR" would heal it, but if it never shows up then it can't be healed - and in that case what happens when the file is appended or truncated for writing?


From my tests of doing a compile in a loop on an replicated mount, if a node goes down then comes back up, the compile will break, doing ls -laR does not fix it once it is broken, You have to make sure the ls -laR happens before ANY OTHER file access, which is impossible in a real environment.

3) File re-opening after a server comes back up

Does the replication get kickstarted by doing a "ls -laR" to initiate self-healing even if a file is still being open and another server has rejoined? Will replication then continue online for currently open files?

Thanks.

Gordan


Maybe a gluster programmer can give better answers than I can. I hope that 
helped. If you have an _immediate_ need for a fuse mountable clustered 
filesystem you might want to look at : 
http://wiki.apache.org/hadoop/MountableHDFS HDFS does replication(+ self heal) 
well. Or im sure these gluster bugs will be fixed soon! :)




On 04/05/2009 21:49, ender wrote:
Also you might want to check here.

http://gluster.org/docs/index.php/Understanding_AFR_Translator#Known_Issues

Gordan Bobic wrote:
Is there a changelog from rc9?
Has the 1st access bug been fixed?

Gordan


_______________________________________________
Gluster-devel mailing list
address@hidden
http://lists.nongnu.org/mailman/listinfo/gluster-devel




_______________________________________________
Gluster-devel mailing list
address@hidden
http://lists.nongnu.org/mailman/listinfo/gluster-devel





reply via email to

[Prev in Thread] Current Thread [Next in Thread]