Re: [xwiki-users] Slow cache sync with JGroups / S3 in an EC2 cluster

2015-05-08 Thread vinc...@massol.net
Hi there,

I don’h have experience using jgroups over s3 but a minute seems really long.

When jgroups is used on a local network it’s instantaneous, there delay is so 
small that it’s unnoticeable.

Are there any recommendations on how to speed up cache-sync time between XWiki 
hosts on a cluster?”

Not much can be doe on the xwiki side here. I think the best bet would be to 
ask on the jgroups mailing lists to see if anyone has experience of jgroups 
over s3.

Also, in a cluster, is the default XWiki cache a distributed cache?”

This is what xwiki is implementing! each instance has a cache and they sync 
through jgroups, making it a distributed cache. The only idea I have would be 
to modify the mechanism somehow so that:
* when the page is saved, jgroups is configured to block till all cluster 
members have received the message
* this could also be implemented by having all members returning another ack 
message to the  sender, and the sender would wait for these messages before 
exiting the save.

However in your case this would mean that any save would take more than 1 
minutes (probably 2+ seen the communication tile you are currently 
experiencing). Thus it won’t be a solution.

You really need to find what’s taking long.

Or is the same cache content replicated across all nodes in a cluster? “

see 
http://extensions.xwiki.org/xwiki/bin/view/Extension/Observation+Module+Remote

no content is transmitted over the network. Only the doc reference so that each 
instance evicts it from its memory cache so that next time the doc is asked 
it’s fetched from the DB.

Hope it helps,
-Vinent Massol
On 8 May 2015 at 03:57:58, Debajit Adhikary (deba...@outlook.com) wrote:

We have deployed XWiki over a two-node EC2 cluster. We are using JGroups over 
S3 for cluster messaging as multicast does not work in EC2.  
I have noticed that if I edit a page on host A, and then refresh the same page 
from host B, it shows stale content — it takes up to a minute for the caches to 
sync.  
Is this something others have observed when running XWiki over multiple hosts? 
(I'm wondering if this is a JGroups/S3 issue)For people who are running XWiki 
in a cluster, how long does it typically take for changes made from one machine 
to show up in another?Are there any recommendations on how to speed up 
cache-sync time between XWiki hosts on a cluster?  
Also, in a cluster, is the default XWiki cache a distributed cache? Or is the 
same cache content replicated across all nodes in a cluster?  
___
users mailing list
users@xwiki.org
http://lists.xwiki.org/mailman/listinfo/users


[xwiki-users] Slow cache sync with JGroups / S3 in an EC2 cluster

2015-05-07 Thread Debajit Adhikary
We have deployed XWiki over a two-node EC2 cluster. We are using JGroups over 
S3 for cluster messaging as multicast does not work in EC2. 
I have noticed that if I edit a page on host A, and then refresh the same page 
from host B, it shows stale content — it takes up to a minute for the caches to 
sync.
Is this something others have observed when running XWiki over multiple hosts? 
(I'm wondering if this is a JGroups/S3 issue)For people who are running XWiki 
in a cluster, how long does it typically take for changes made from one machine 
to show up in another?Are there any recommendations on how to speed up 
cache-sync time between XWiki hosts on a cluster?
Also, in a cluster, is the default XWiki cache a distributed cache? Or is the 
same cache content replicated across all nodes in a cluster?

___
users mailing list
users@xwiki.org
http://lists.xwiki.org/mailman/listinfo/users