ok, let me more clarify this topic since I got one more email about not
using memcache for locking.
First of all, I know memcached is not reliable by design but if you
visit http://en.wikipedia.org/wiki/Memcached you will see that I'm not
the only/first person to try this.( there is a link to
IPC::Lock::Memcached - fast locking via memcached).
For more information about my software, the locking mechanism is also
backed up by a MVCC implementation in db layer. In other words, I
already have a backup mechanism if memcached fails not to loose data
integrity, but the cost of rolling back a bunch of transactions is not
small, so I dont want to see MVCC exceptions hanging around.
In addition, I need to guarantee that at least one memcached server is
alive not to freeze my system.(e.g. if nobody can obtain lock, nobody
can take an action).
Now comes the case of saving locks when a server fails. In that case, we
may loose ~ 10K locks which will cause ~ 7K operation rollbacks. What I
think is, with a memcache master slave implementation, this number can
be decreased to a ommittable number.
Of course there is another solution to this problem, basically depending
on db layer MVCC implementation if memcache fails, but this will be a
highly coupling between these two, which wont be preferable by design.
Anyway, to clarify, any master-slave like replication of memcached may
be helpful for my case.
Henrik Schröder wrote:
The best practice is to not use memached at all for this. It is a
cache, if you store something in it, you are in no way guaranteed that
it will be there the next time you ask for it. It is temporary, not
permanent, and as such an extremely bad match for a distributed lock
where it is absolutely critical that you do not lose data.
/Henrik
On Thu, Mar 12, 2009 at 19:30, Yigit Boyar <[email protected]
<mailto:[email protected]>> wrote:
hi all,
I'm developing a software where memcached is also used as locking
server.
The problem is that, i wanna make this fail safe; which basically
means if the memcached server goes down, another one starts and
the locks are transformed. This basically handled by keeping two
instances up, one as master and if the master dies, slave becomes
master and an new slave comes live.
This is my guess, but I'm looking for some good best practices not
to invent wheel from scratch.
So is there any link you can provide where a best practice for
memcached based locking is present ?
btw. i'm using php in a distributed environment.
thnks.
