From mboxrd@z Thu Jan  1 00:00:00 1970
From: Junxiao Bi <junxiao.bi@oracle.com>
Date: Tue, 21 Jun 2016 17:03:04 +0800
Subject: [Ocfs2-devel] [PATCH v2] ocfs2: improve recovery performance
In-Reply-To: <5767CF2C020000F90003CB5F@prv-mh.provo.novell.com>
References: <1466155682-24656-1-git-send-email-junxiao.bi@oracle.com>
	<5763C644.9060104@huawei.com>
	<5767CF2C020000F90003CB5F@prv-mh.provo.novell.com>
Message-ID: <576902C8.30503@oracle.com>
List-Id: <ocfs2-devel.oss.oracle.com>
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
To: ocfs2-devel@oss.oracle.com

Hi Gang,

On 06/20/2016 11:10 AM, Gang He wrote:
> Hello Junxiao,
> 
> I think this change will bring a performance improvement, but from the function comments
> /*
>  * JBD Might read a cached version of another nodes journal file. We
>  * don't want this as this file changes often and we get no
>  * notification on those changes. The only way to be sure that we've
>  * got the most up to date version of those blocks then is to force
>  * read them off disk. Just searching through the buffer cache won't
>  * work as there may be pages backing this file which are still marked
>  * up to date. We know things can't change on this file underneath us
>  * as we have the lock by now :)
>  */
> static int ocfs2_force_read_journal(struct inode *inode)
> 
> Did we consider this potential risk behind this patch? I am not familiar with this part code, 
> I want to know if there is any sync mechanism to make sure the block cache for another node journal file is really the latest data?  
I don't see that is needed, because those stale info will not be used
except journal replay.

Thanks,
Junxiao.
> 
> 
> 
> Thanks
> Gang 
> 
> 
>>>>
>> On 2016/6/17 17:28, Junxiao Bi wrote:
>>> Journal replay will be run when do recovery for a dead node,
>>> to avoid the stale cache impact, all blocks of dead node's
>>> journal inode were reload from disk. This hurts the performance,
>>> check whether one block is cached before reload it can improve
>>> a lot performance. In my test env, the time doing recovery was
>>> improved from 120s to 1s.
>>>
>>> Signed-off-by: Junxiao Bi <junxiao.bi@oracle.com>
>> Looks good to me. And it indeed has performance improvement from my
>> test.
>> Reviewed-by: Joseph Qi <joseph.qi@huawei.com>
>>
>>> ---
>>>  fs/ocfs2/journal.c |   41 ++++++++++++++++++++++-------------------
>>>  1 file changed, 22 insertions(+), 19 deletions(-)
>>>
>>> diff --git a/fs/ocfs2/journal.c b/fs/ocfs2/journal.c
>>> index e607419cdfa4..bc0e21e8a674 100644
>>> --- a/fs/ocfs2/journal.c
>>> +++ b/fs/ocfs2/journal.c
>>> @@ -1159,10 +1159,8 @@ static int ocfs2_force_read_journal(struct inode 
>> *inode)
>>>  	int status = 0;
>>>  	int i;
>>>  	u64 v_blkno, p_blkno, p_blocks, num_blocks;
>>> -#define CONCURRENT_JOURNAL_FILL 32ULL
>>> -	struct buffer_head *bhs[CONCURRENT_JOURNAL_FILL];
>>> -
>>> -	memset(bhs, 0, sizeof(struct buffer_head *) * CONCURRENT_JOURNAL_FILL);
>>> +	struct buffer_head *bh = NULL;
>>> +	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
>>>  
>>>  	num_blocks = ocfs2_blocks_for_bytes(inode->i_sb, i_size_read(inode));
>>>  	v_blkno = 0;
>>> @@ -1174,29 +1172,34 @@ static int ocfs2_force_read_journal(struct inode 
>> *inode)
>>>  			goto bail;
>>>  		}
>>>  
>>> -		if (p_blocks > CONCURRENT_JOURNAL_FILL)
>>> -			p_blocks = CONCURRENT_JOURNAL_FILL;
>>> +		for (i = 0; i < p_blocks; i++) {
>>> +			bh = __find_get_block(osb->sb->s_bdev, p_blkno,
>>> +					osb->sb->s_blocksize);
>>> +			/* block not cached. */
>>> +			if (!bh) {
>>> +				p_blkno++;
>>> +				continue;
>>> +			}
>>>  
>>> -		/* We are reading journal data which should not
>>> -		 * be put in the uptodate cache */
>>> -		status = ocfs2_read_blocks_sync(OCFS2_SB(inode->i_sb),
>>> -						p_blkno, p_blocks, bhs);
>>> -		if (status < 0) {
>>> -			mlog_errno(status);
>>> -			goto bail;
>>> -		}
>>> +			brelse(bh);
>>> +			bh = NULL;
>>> +			/* We are reading journal data which should not
>>> +			 * be put in the uptodate cache.
>>> +			 */
>>> +			status = ocfs2_read_blocks_sync(osb, p_blkno, 1, &bh);
>>> +			if (status < 0) {
>>> +				mlog_errno(status);
>>> +				goto bail;
>>> +			}
>>>  
>>> -		for(i = 0; i < p_blocks; i++) {
>>> -			brelse(bhs[i]);
>>> -			bhs[i] = NULL;
>>> +			brelse(bh);
>>> +			bh = NULL;
>>>  		}
>>>  
>>>  		v_blkno += p_blocks;
>>>  	}
>>>  
>>>  bail:
>>> -	for(i = 0; i < CONCURRENT_JOURNAL_FILL; i++)
>>> -		brelse(bhs[i]);
>>>  	return status;
>>>  }
>>>  
>>>
>>
>>
>>
>> _______________________________________________
>> Ocfs2-devel mailing list
>> Ocfs2-devel at oss.oracle.com 
>> https://oss.oracle.com/mailman/listinfo/ocfs2-devel
>