Mailing-List: contact dev-help@hbase.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@hbase.apache.org
Received-SPF: pass (nike.apache.org: domain of ryanobjc@gmail.com designates
 209.85.210.169 as permitted sender)
DomainKey-Signature: a=rsa-sha1; c=nofws;
        d=gmail.com; s=gamma;
        h=mime-version:in-reply-to:references:date:message-id:subject:from:to
         :cc:content-type:content-transfer-encoding;
        b=F4P1z8N0a1KtME0BUv2He8lwdgjuDUYOw6Io6SuR37/RqEI3vpOTAdUkSu9brek5Hg
         /dQ7FF24FAbTup+VHuMa4ULMnZtXPaTF1toaisNWwM1ymKP8/ZGd0Lv40pG2OkRl8t4a
         5tBvvItH31OU4g71u5npku8lqlJr5aIjQcPbU=
MIME-Version: 1.0
In-Reply-To: <AANLkTimH_WKx6cFYjWNJN1yhuuUkx+C5WkYN4zAH0tda@mail.gmail.com>
References: <AANLkTimK4LsnAEhi1ZWuQ+sSv3uUrme_4AGxZ-Aq0mFP@mail.gmail.com>
	<AANLkTimH_WKx6cFYjWNJN1yhuuUkx+C5WkYN4zAH0tda@mail.gmail.com>
Date: Tue, 1 Mar 2011 12:08:31 -0800
Message-ID: <AANLkTi=25e0fn-SxvDwhyq2iTGq+aR+=_G0zC0d9FZAu@mail.gmail.com>
Subject: Re: Coprocessor tax?
From: Ryan Rawson <ryanobjc@gmail.com>
To: dev@hbase.apache.org
Cc: Gary Helmling <ghelmling@gmail.com>
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable

My own profiling shows that a read write lock can be up to 3-6% of the
CPU budget in our put/get query path.  Adding another one if not
necessary would probably not be good.

In fact in the region coprocessor the only thing the write lock is
used for is the preClose and postClose, but looking in the
implementation of those methods I don't really get why this is
necessary.  The write lock ensures single thread access, but there is
nothing that prevents other threads from calling other methods AFTER
the postClose?

-ryan

On Tue, Mar 1, 2011 at 12:02 PM, Gary Helmling <ghelmling@gmail.com> wrote:
> All the CoprocessorHost invocations should be wrapped in "if (cpHost !=3D
> null)". =A0We could just added an extra check for whether any coprocessor=
s are
> loaded -- "if (cpHost !=3D null && cpHost.isActive())", something like th=
at?
> =A0Or the CoprocessorHost methods could do this checking internally.
>
> Either way should be relatively easy to bypass the lock acquisition. =A0I=
s
> there much overhead to acquiring a read lock if the write lock is never
> taken though? (just wondering)
>
>
>
> On Tue, Mar 1, 2011 at 11:51 AM, Stack <stack@duboce.net> wrote:
>
>> So, I'm debugging something else but thread dumping I see a bunch of thi=
s:
>>
>>
>> "IPC Server handler 6 on 61020" daemon prio=3D10 tid=3D0x00000000422d280=
0
>> nid=3D0x7714 runnable [0x00007f1c5acea000]
>> =A0 java.lang.Thread.State: RUNNABLE
>> =A0 =A0 =A0 =A0at
>> java.util.concurrent.locks.ReentrantReadWriteLock$Sync.fullTryAcquireSha=
red(ReentrantReadWriteLock.java:434)
>> =A0 =A0 =A0 =A0at
>> java.util.concurrent.locks.ReentrantReadWriteLock$Sync.tryAcquireShared(=
ReentrantReadWriteLock.java:404)
>> =A0 =A0 =A0 =A0at
>> java.util.concurrent.locks.AbstractQueuedSynchronizer.acquireShared(Abst=
ractQueuedSynchronizer.java:1260)
>> =A0 =A0 =A0 =A0at
>> java.util.concurrent.locks.ReentrantReadWriteLock$ReadLock.lock(Reentran=
tReadWriteLock.java:594)
>> =A0 =A0 =A0 =A0at
>> org.apache.hadoop.hbase.regionserver.RegionCoprocessorHost.prePut(Region=
CoprocessorHost.java:532)
>> =A0 =A0 =A0 =A0at
>> org.apache.hadoop.hbase.regionserver.HRegion.doMiniBatchPut(HRegion.java=
:1476)
>> =A0 =A0 =A0 =A0at
>> org.apache.hadoop.hbase.regionserver.HRegion.put(HRegion.java:1454)
>> =A0 =A0 =A0 =A0at
>> org.apache.hadoop.hbase.regionserver.HRegionServer.multi(HRegionServer.j=
ava:2652)
>> =A0 =A0 =A0 =A0at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Me=
thod)
>> =A0 =A0 =A0 =A0at
>> sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.jav=
a:39)
>> =A0 =A0 =A0 =A0at
>> sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessor=
Impl.java:25)
>> =A0 =A0 =A0 =A0at java.lang.reflect.Method.invoke(Method.java:597)
>> =A0 =A0 =A0 =A0at
>> org.apache.hadoop.hbase.ipc.WritableRpcEngine$Server.call(WritableRpcEng=
ine.java:309)
>> =A0 =A0 =A0 =A0at
>> org.apache.hadoop.hbase.ipc.HBaseServer$Handler.run(HBaseServer.java:106=
0)
>>
>>
>> Do others? =A0I don't have any CPs loaded. =A0I'm wondering if we can do
>> more to just avoid the CP codepath if no CPs loaded.
>>
>> St.Ack
>>
>