On Mon, Nov 11, 2013 at 1:09 PM, Varun Sharma <[EMAIL PROTECTED]> wrote:
> I am looking through the code - it seems like HConnectionImplementation
> keeps rpcTimeout as a class variable. I thought there is a single
> HConnection object and there is a factory of some Configuration Attributes
> (including rpc timeout) -> HConnections. Am I missing something - I dint
> see a threadlocal variable anywhere at least in the 0.94 code base ?
That is right. The rpcTimeout is passed when we set up the
proxy/connection and then again as timeout waiting on root region. Inside
in the proxy, there is a thread local also named rpcTimeout (see head of
On each call, we pass in the rpcTimeout. If it is different than default,
we actually make a new connection (See ConnectionId setup).
HTable operation timeout is passed to the call so looks like it is rpc
It is tough to follow. If still makes no sense, just say and I'll do a
> Also, one more thing - I see a client operation timeout in HTable beign
> used as well - is that different from the rpc timeout.
> Thanks !
> On Mon, Nov 11, 2013 at 12:26 PM, Stack <[EMAIL PROTECTED]> wrote:
> > On Mon, Nov 11, 2013 at 12:11 PM, Varun Sharma <[EMAIL PROTECTED]>
> > wrote:
> > > Hi,
> > >
> > > Can hbase rpc timeout be changed across different HBase rpc calls for
> > HBase
> > > 0.94. From the code, it looks like this is not possible ? I am
> > if
> > > there is a way to fix this ?
> > >
> > I'm guessing you have already dug in and noticed rpcTimeout is a thread
> > local that I'd think you'd be able to get access too w/ some
> > but this doesn't work for you?
> > St.Ack