[Gluster-users] 回复： GlusterFS use question

[Joe Julian]

On 07/27/2014 07:47 PM, yan wrote:
> hi,*david.zhang700*‍
> now configure
> gluster volume set gfs network.ping-timeout 1
Bad idea. Now even a one second delay will cause a disconnect. Under 
load, it's entirely possible to have a 1 second delay reconnecting due 
to open fds, locks, etc. That delay will cause the client to drop the 
connection. Then it will reconnect, have the same delay, and drop the 
connection.

Why is 42 seconds too long? Is your network so bad that you lose 
connection so often that a more stable cluster is a bad thing?

> ‍
> client  reconnect ‍shorten the time‍
> but ，have a question
> client down,start client ,data Not immediately synchronization‍
> how configure immediately synchronization?‍
>

That just seems impractical with regard to cpu and network usage. I 
would much rather have my servers servicing client requests than 
spending all their time spinning to see if an absent server has returned.

It seems that it checks every 3 seconds now to see if it's returned. 
That goes back to my previous statement. If your network is so unstable 
that 3 seconds to start self-heal is excessive, perhaps you might look 
at a more stable network.

That said, feel free to open an enhancement request on bugzilla to allow 
the user to tune the reconnect interval.

> ------------------ 原始邮 件 ------------------
> *发件人:* "david.zhang700";<david.zhang700 at gmail.com>;
> *发送时间:* 2014年7月25日(星期五) 晚上7:09
> *收件人:* "yan"<yanlk at foxmail.com>; 
> "gluster-users"<gluster-users at gluster.org>;
> *主题:* Re: [Gluster-users] GlusterFS use question
>
> gluster volume set volume_name network.ping-timeout seconds_you_want
> default is 42s
> David
> *From:* yan <mailto:yanlk at foxmail.com>
> *Sent:* Friday, July 25, 2014 4:51 PM
> *To:* gluster-users <mailto:gluster-users at gluster.org>
> *Subject:* [Gluster-users] GlusterFS use question
> hi ,
> i am have a question? ‍
> server:
> 10.10.10.51
> 10.10.10.52‍
> 10.10.10.53‍
> 10.10.10.54‍
> client:
> 10.10.10.55
> server configure:
> gluster mode replica
> ‍
> eg.
> gluster volume create gfs replica 4 transport tcp 
> 10.10.10.51:/data/file1 10.10.10.52:/data/file1 
> 10.10.10.53:/data/file1 10.10.10.54:/data/file1 force
> client :
> mount -t glusterfs 10.10.10.51:/gfs /mnt/gfs‍
> Now ,i am have a question
> if one of four server down; eg 10.10.10.51 down.
> client is disconnect ,I am need wait 90seconds, client is autoconnect, 
> clinet is ok
>
> How to configure the serverto shorten the time.‍
>
> ------------------------------------------------------------------------
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://supercolony.gluster.org/mailman/listinfo/gluster-users
>
>
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://supercolony.gluster.org/mailman/listinfo/gluster-users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://supercolony.gluster.org/pipermail/gluster-users/attachments/20140727/7f98da1e/attachment.html>