Mailing-List: contact dev-help@hbase.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@hbase.apache.org
Received-SPF: pass (athena.apache.org: domain of yuzhihong@gmail.com
 designates 209.85.220.47 as permitted sender)
References: 
 <CAPQV63VyurYdFTJqJF8izkhNMVMLwc5amXfKrD8FzQkXaP34PQ@mail.gmail.com>
 <CANZa=GtjYZizjCGr7L07UStPpnwFrBGev_SwWUN8OuToj7r5WQ@mail.gmail.com>
 <CALte62zrAcCMu+VSw9jh1Ob4nQ-zy5XEBXSxguGOC6RxwkAw8Q@mail.gmail.com>
 <CAPQV63XJrX+OdQvSKa0Y8QaPb0ce2g-S46-Wnr24oqQ-ivbPcg@mail.gmail.com>
Mime-Version: 1.0 (1.0)
In-Reply-To: 
 <CAPQV63XJrX+OdQvSKa0Y8QaPb0ce2g-S46-Wnr24oqQ-ivbPcg@mail.gmail.com>
Content-Type: text/plain;
	charset=utf-8
Content-Transfer-Encoding: quoted-printable
Message-Id: <3B2631CD-5068-4228-899A-EFDE8B98F934@gmail.com>
Cc: "dev@hbase.apache.org" <dev@hbase.apache.org>
From: Ted Yu <yuzhihong@gmail.com>
Subject: Re: Corrupted log file
Date: Sun, 5 May 2013 06:01:30 -0700
To: "dev@hbase.apache.org" <dev@hbase.apache.org>

Looks like breaking the loop is better choice.=20

Cheers

On May 5, 2013, at 4:38 AM, Jean-Marc Spaggiari <jean-marc@spaggiari.org> wr=
ote:

> Ok, I will open a JIRA for that later today...
>=20
> On the RS side, should we break the loop? Or kill the server?  Because not=

> being able to read the log might end with rs inconsistencies?
>=20
> JM
> Le 1 mai 2013 14:51, "Ted Yu" <yuzhihong@gmail.com> a =C3=A9crit :
>=20
>> Ideally HBCK should sideline corrupted log file so that region server can=

>> start.
>>=20
>> Cheers
>>=20
>> On Wed, May 1, 2013 at 11:48 AM, Nick Dimiduk <ndimiduk@gmail.com> wrote:=

>>=20
>>> Detecting the condition, printing the warning, and breaking the loop
>> sounds
>>> like an urgent bandaid solution to me.
>>>=20
>>> -n
>>>=20
>>> On Tue, Apr 30, 2013 at 11:51 AM, Jean-Marc Spaggiari <
>>> jean-marc@spaggiari.org> wrote:
>>>=20
>>>> When a log file (into /hbase/.logs) is corrupted, HBase is not able to
>>>> start because it tries to read it again and again.
>>>>=20
>>>> Also, there is nothing into HBCK to detect that.
>>>>=20
>>>> Should we have something to check that? Like in hbck, we can simply try=

>>> to
>>>> open the log file, and read it? To report the warning?
>>>>=20
>>>> JM
>>=20