james-server-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ioan Eugen Stan <stan.ieu...@gmail.com>
Subject Re: hbase-mailbox alpha release
Date Tue, 09 Aug 2011 16:13:15 GMT
2011/8/9 Robert Burrell Donkin <robertburrelldonkin@gmail.com>:
> On Tue, Aug 9, 2011 at 3:17 PM, Ioan Eugen Stan <stan.ieugen@gmail.com> wrote:
>> 2011/8/9 Robert Burrell Donkin <robertburrelldonkin@gmail.com>:
>>> On Mon, Aug 8, 2011 at 9:57 PM, Ioan Eugen Stan <stan.ieugen@gmail.com>
>>>> I fixed the exception in the test (forgot to update it when I made the
>>>> change). I also fixed some other issues.
>>>> Now all tests should pass, except for HBaseMailboxManager which runs
>>>> indefinitely. I tried to profile the test case in NetBeans but failed.
>>>> I have an idea of what's going on, but not being able to profile the
>>>> application will probably take more time.
>>> One tactic I sometimes use with type of problem is to add more
>>> specific tests. Another is to selectively dummy.
>> I have tracked down the problem. It is in HBaseMailboxMapper.save() method. [1].
>> I have put some System.out.print statements to see what happens and it
>> seems that reading from the InputStream blocks.
>> This is strange. I will try to see what is causing this. Do the
>> messages created in the test suite have content equal to the size?
>>            while ((b = message.getBodyContent().read()) != -1) {
>>                out.write(b);
>>            }
>>            out.close();
>>            out = new
>> ChunkOutputStream(sessionMapperFactory.getClusterConfiguration(),
>> messageRowKey(message), MAX_COLUMN_SIZE);
>>            while ((b = message.getHeaderContent().read()) != -1) {
>>                out.write(b);
>>            }
> I've just had a hard power failure and I'm recovering my main
> development box now :-/
> It might be tomorrow before I can get back to you but hopefully
> someone else will be able to jump in before then...
> Robert

I hope everything will be ok. Regarding the above problem, it's
solved: I moved message.get*Content outside the wile loop so I will
keep only one reference and read from that stream. I also used a
BufferedInputStream to speed up thing.

Unfortunately, save is not done yet. It seems to use a lot of memory.
If I eliminate save() by returning SimpleMessageMetadate from the
start, the test is green so the problem lies here.

I will keep you posted. Good luck with the recovery process.


Ioan Eugen Stan

To unsubscribe, e-mail: server-dev-unsubscribe@james.apache.org
For additional commands, e-mail: server-dev-help@james.apache.org

View raw message