| > In our DMQ program, we use PDEL_MODE_WF_DQF delivery mode and PDEL_UMA_SAF
>uma mode. After completion of pams_put_msg, if the disk is crashed, which DQF
>file resides, or DQF file is crashed, what happen ?
> In this case, pls explane MRS server's action and how can I recover message?
(Lisa may expand on this)
What the MRS Server does on startup is first sort the files within
each DQF and SAF file set. This is done because file sequence numbers
can be reused and end up being sequentialy out of order. This sorting
requires MRS to find the first msg in each file. Unless you have
suppressed the counting od msgs in the journal, the MRS scan the
complete file. After this the MRS Server is not ready to process
normal traffic.
With the above said, should the MRS Server be unable to read correctly
any file it will attempt to bypass the msg (of course logging an error)
or if it can't read the file it will rename the file to .DERR or .SERR
in order to save it for the possibility of recovery later on.
There is only so much MRS can do. If you want to survive disk
failure then use some level of RAID to eliminate the single points of
failures.
Marty
|
| Hi, Marty.
I understand the MRS sever's doing on startup. But after startup MRS server,
if the DQF file crashed suddenly (disk is normal state), Is there any method
to recover DQF file or message immediately ?
We are free from disk failure because we use RAID1(volume shadowing).
And would you tell me the point of DMQ V3.2B kit location ?
Thank you sincerely,
- Hoon
|
| If processing is halted suddenly on the MRS_SERVER process due to
a hardware or operating system crash, then normal file system
rundown may leave the DQF with an incomplete last record. This is
always detected when MRS restarts. Since notification of arrival
at the DQF is only given when MRS is told by the OS that the record
is on the disk, there is no information lost in this scenerio.
On the other hand, there is NOTHING the software can do for handling
file system corruption. Protection from this (rare) form of failure
must by redundant hardware like Marty suggests. MRS checks record
signatures for file integrity but does not go further than that.
|