How to know what caused Data loss ??

[Kiran Kumar]

I am working on a application that uses two memcache servers configured as a cluster . I am currently in a stage of testing the application , for this sometimes i manually make one instance of MEmcache down .

Sometimes our network goes Up/down and as a case it sometimes doen't recieve the data , i am in a situation where i am confused whether the data loss is due to Network issues /or making a memcace instance down .

For your information , i am using xmemcached 1.3.8 , ketamamemcachesessionlocator and repcached 2.3 for replication with

Failure Mode off in client .

So i am pretty much sure that there will be no Data loss at all (Correct me if i am wrong )

My question is that , is there any situation by which i can know that this loss is due to what cause ??

[dennis zhuang]

Hi,memcached is juat a cache in memory,it will lost all data after restart.

I don't know repcache replicates the data between memacheds synchronously.If it doesn't, then maybe some data will be lost when one memcached was crashed.

If the set method in MemcachedClient returns true,that means xmemcached have stored the data to memcached,but the data may be lost by LRU replcament or crash.

2012/11/20 Kiran Kumar <krn...@gmail.com>

--
You received this message because you are subscribed to the Google Groups "xmemcached" group.
To view this discussion on the web visit https://groups.google.com/d/msg/xmemcached/-/ryKK2eXxb0wJ.
To post to this group, send email to xmemc...@googlegroups.com.
To unsubscribe from this group, send email to xmemcached+...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/xmemcached?hl=en.

--
庄晓丹
Email:        killm...@gmail.com xzh...@avos.com
Site:           http://fnil.net
Twitter:      @killme2008

[Kiran]

Thanks Dennis , there is no chance of LRU replacement as the telnet stats commands show that the stats bytes is very less when compared to the stats limitmaxbytes .

So there is no chance of LRU replacement here i guess so .

And about the  Data loss hapened due to memcached crash , i cannot rule out the issue that the network is very slow at that time and the stable version of that application which is running  in production is also behaving the same  , and i observed that the Data is coming on to the screen within  5-6  minutes of time , This is what i wanted to verify actually .

That point you mentioned "set method in MemcachedClient returns true" will be veryhelpful for debugging .

Thanks agin Dennis , for being so  helpful