Skip to contentSkip to conversation

Disk full and writting to pre-allocated area on ReiserFS

DISCUSSION Peter Zaitsev 4 participants 4 messages Oct 15, 2004 – Oct 18, 2004

lore
Share & tools
lore
#0 lore raw

Hi,

I'm running SuSE 9.1 Kernel 2.6.5-7.108-default
But I would guess it applies to large variety of platforms as we have customers reporting the same problem.

I'm using reiserfs:
/dev/md0 on /data type reiserfs (rw,noatime,notail,data=writeback)

The problem is in case of disk full condition, "Disk full" error is being reported even if write happens to Pre-Allocated area, in my case to Innodb recovery log files.

This is very unfortunate as in such case Innodb has no way but to terminate database server. These logs are specially pre-allocated so one would not run in such condition.

Question: Is there any way to avoid this problem with Reiserfs ?

Signature · 2 lines
Peter Zaitsev, Senior Support Engineer
MySQL AB, www.mysql.com
#1 lore raw
Quoted 8 lines · Peter Zaitsev

Hi,

I'm running SuSE 9.1 Kernel 2.6.5-7.108-default But I would guess it applies to large variety of platforms as we have customers reporting the same problem.

I'm using reiserfs:
/dev/md0 on /data type reiserfs (rw,noatime,notail,data=writeback)

Quoted 12 lines

The problem is in case of disk full condition, "Disk full" error is being reported even if write happens to Pre-Allocated area, in my case to Innodb recovery log files.

This is very unfortunate as in such case Innodb has no way but to
terminate database server. These logs are specially pre-allocated so one would not run in such condition.

Question: Is there any way to avoid this problem with Reiserfs ?

vs or chris or jeff, can you comment?

#2 lore raw

Hello

Could someone please forward this mail to [email protected] (for unknown reasons I can not send anything to some addresses).

Quoted 15 lines · Hans Reiser
Quoted 13 lines · Peter Zaitsev

Hi,

I'm running SuSE 9.1 Kernel 2.6.5-7.108-default
But I would guess it applies to large variety of platforms as we have customers reporting the same problem.

I'm using reiserfs:
/dev/md0 on /data type reiserfs (rw,noatime,notail,data=writeback)

The problem is in case of disk full condition, "Disk full" error is being reported even if write happens to Pre-Allocated area, in my case to Innodb recovery log files.

Quoted 6 lines
Quoted 6 lines

This is very unfortunate as in such case Innodb has no way but to terminate database server. These logs are specially pre-allocated so one would not run in such condition.

Question: Is there any way to avoid this problem with Reiserfs ?

Would you please check whether attached patch fixes the problem?

Quoted 5 lines
Quoted 3 lines

vs or chris or jeff, can you comment?

This looks like a bug in reiserfs_file_write. It does not distinguish whether new disk space is to be allocate.

file.c renamed from file.c~ +2 -0
@@ -1137,6 +1137,8 @@
 	return result;
     }
 
+    return generic_file_write(file, buf, count, ppos);
+
     if ( unlikely((ssize_t) count < 0 ))
         return -EINVAL;
 
#3 lore raw
Quoted 2 lines · Vladimir Saveliev

Could someone please forward this mail to [email protected] (for unknown reasons I can not send anything to some addresses).

Forwarded.

Signature · 1 line
mjt