HP-UX HB v13.00 Ch-14 - JFS

HP-UX Handbook – Rev 13.00 Page 24 (of 47)
Chapter 14 Journaled File System (JFS)
October 29, 2013
Depending on the speed of the storage device, if you lower write_throttle, user write
performance may suffer, but the number of dirty buffers is limited, so sync operations complete
much faster.
Because lowering write_throttle may in some cases delay write requests (for example, lowering
write_throttle may increase the file disk queue to the max_diskq value, delaying user writes until
the disk queue decreases), it is advisable not to change the value of write_throttle unless your
system has a combination of large physical memory and slow storage devices.
Troubleshooting File System Problems
There are three different ways how a file system reacts to problems:
 marks an inode bad
this happens if an inode update or a directory- block update fails. In this case any attempt
to access the data in the file or change the inode will fail.
Inodes can be marked bad if an inode update or a directory-block update fails.
In these types of failures, the file system doesn't know what information is on the disk,
and considers all the information that it finds to be invalid. After an inode is marked bad,
the kernel still permits access to the file name, but any attempt to access the data in
the file or change the inode fails.
 disables transactions
this happens if the file system detects an error while writing to the intent log. In this case
block or inode frees or allocation, structural changes, directory entry changes will fail.
If the file system detects an error while writing the intent log, it disables transactions.
After transactions are disabled, the files in the file system can still be read or written, but
no block or inode frees or allocations, structural changes, directory entry changes, or
other changes to metadata are allowed.
 disables the file system
this happens if a superblock update fails.
If an error occurs that compromises the integrity of the file system, VxFS disables itself.
If the intent log fails or an inode-list error occurs, the super-block is ordinarily updated
(setting the VX_FULLFSCK flag) so that the next fsck does a full structural check. If this
super-block update fails, any further changes to the file system can cause inconsistencies
that are undetectable by the intent log replay.
To avoid this situation, the file system disables itself.
Anytime the file system encounters a problem it will write an error message to the syslog.log In
nearly every case the file system must be umounted and a fsck has to be performed.
The main reason for file system errors result from hardware failures, hence it is recommended to