zookeeper-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Prabhjot Bharaj <prabhbha...@gmail.com>
Subject Re: Zookeeper leader not getting lock
Date Fri, 29 May 2015 06:48:01 GMT
Hi Chris,

Thanks for your email. Yes, you understood the question very well.
I'm using the standard api in python through the zookeeper.acquire call

I was a little confused but, thanks to your email and some reading that I
did after posting this question, I understood it.
The client can be any machine, not just the server nodes and because of
that I had got confused.



Regards,
Prabhjot


On Fri, May 29, 2015 at 12:28 AM, Chris Nauroth <cnauroth@hortonworks.com>
wrote:

> Hello prabcs,
>
> When you mention that "all 3 nodes try to acquire the lock", does this
> mean that all 3 nodes run a ZooKeeper client process that tries to acquire
> a shared lock, using something like Curator or the documented locking
> recipes?
>
> http://zookeeper.apache.org/doc/r3.4.6/recipes.html#sc_recipes_Locks
>
>
> If so, then it doesn't matter whether the client process is co-located on
> a machine running a leader or a follower.  Each of the 3 nodes triggers
> the script independently via cron.  Therefore, the order in which these
> processes start and attempt to acquire the lock is non-deterministic.  The
> script co-located on the leader node might win and get the lock, or the
> script co-located on one of the follower nodes might win instead.
>
> I hope this helps.  If I misunderstood the question, please let me know.
>
> --Chris Nauroth
>
>
>
>
> On 5/28/15, 11:40 AM, "Prabhjot Bharaj" <prabhbharaj@gmail.com> wrote:
>
> >Hi Folks,
> >
> >
> >I have a 3 node zookeeper cluster. While working on it, I've come to
> >notice
> >that sometimes that lock gets acquired by the follower and the leader is
> >unable to acquire the lock. I'm able to see this clearly as per the
> >script's output log. This script runs as a cron on all the 3 nodes and all
> >3 nodes try to acquire the lock
> >
> >After my script runs on all the nodes, I can see by running echo stat | nc
> >localhost 2181 on all machines that the machine which has got the lock and
> >ran the script successfully is a follower and not a leader.
> >
> >I'm thinking that the follower should not be able to get a lock. Am I
> >thinking wrong somewhere ? Please point me to the correct direction.
> >
> >Thanks,
> >
> >prabcs
>
>


-- 
---------------------------------------------------------
"There are only 10 types of people in the world: Those who understand
binary, and those who don't"

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message