HI,

 Last MFNetDisk recovery was better then before but not good enough.
 I rewrite all the error recovery to be automatically after fixing the TCP
connection.
I made my test to my CIOD (PAV) and regular 3390 MFNetDisk disks and tape,
but I need more feedback and tests.
I put the new code and documentation in my site.
I need feedback. If someone have deep knowledge in IO and can suggest me how
to do it better if possible, you welcome.
I add some pages from my MPCDOC.PDF documentation about the new TCP (or any
error) recovery error.
So, please download my new code using CODE_REQ and try the error recovery.
The code is from April 2, 2011.

Shai

*

MFNetDisk TCP connection error
*

MFNetDisk TCP connection error happens when the TCP connection from MF to PC
is in error. No IO can be sent to PC and the disk and tape emulation can not
work.

Fixing the TCP connection will fix the problem automatically, but MFNetDisk
let you to handle the connection error also manually depend in your case.

To explain the MFNetDisk TCP error handler, I used a real case of TCP
connection error and the recovery process and the messages from MFNetDisk
and MF which will be displayed in case of the error.

All the messages in this example displayed, No messages deleted, everything
is true and real case.

Why I do it? Because in real case you will see many messages and by looking
in my example you will be ready for these messages and not be scared.

In our example I used device 300-301 which are disk 3390 emulation and
320-321 CIOD (Like IBM PAV) disk 3390 emulation.

Be aware that during the recovery process, when the MF sends message of
missing interrupts, do not try to rerun the VARY ONLINE command. This
command may cause some problem because MFNetDisk is virtual devices. Better
to let MFNetDisk do the job.

Anyway, in MFNetDisk code I handle any OS command include VARY if you will
send it, but be kind and let MFNetDisk handle it OK in full control before
doing so.

At 10:44:25.48 you ran MVS command to check the IO connection and verify the
TCP connection status of MFNetDisk.

Until 10:44:49.64 MF IO send messages error messages that to IO end status
is missing. MFNetDisk will try to keep the IO and disabling cancel the IO by
the MF IOS.

After 10:44:49.64 MFNetDisk find out the there is TCP connection problems
after timeout.

At 10:44:49.77 MFNetDisk will play with TCP. MFNetDisk will close the
connection and will try another connection to the PC.

At 10:45:19.99 MFNetDisk determines that the connection is broken.

MFNetDisk will react according to setting value of TPIOE.

MFNetDisk have two options:

1. TPIOE=Y, Purge the pending IO at timeout.

2. TPIOE=N, Continue to retry the IO until the TCP connection will be
recovered.

Option 1 is risky because the IO will be purge and recovering the IO will
not be done. If the user knows for sure that no files and disks need to be
recovered he can use this option. In this case the IOS message will be
stopped and the pending IO will be purged with error status.

The default is option 2 and the recommended option, which is to continue to
retry. The reason is that after connection failed, MF will recover the IO
after the TCP connection recovered. The recovery can be cleaning the VTOC
and more.

In our example option 2 is active and MF will continue to retry until the
connection will be recovered.

At 10:46:08.68 The TCP connection recovered. No needs to run any modify
command. The connection will be recovered automatically.

Another example is to use option 1, which is to purge the IO.

This is risky option because the MF IOS will not recover the IO, but
sometime the user will do it. Example, if you install MFNetDisk for the
first time and you want to check the connection by using VARIES ONLINE
command or DS P, UUU, 1 command. Both commands do not allocated file or
using the VTOC. So there is no danger to use option 1 which is to purge the
IO without fixing the TCP connection and the IO recover is not needed.

In our new example the option 2 is currently active.

At 10:47:22.57 I disable the TCP connection to PC and run few MF console
commands.

At 10:48:22.93 The MFNetDisk timeout reached. Message MPC050I (option 2)
indicate that retry will continue, but you can change the default by using
the F MPC, TPIOE=Y which will set the TPIOE to Y meaning purging the IO in
case of timeout.

At 10:48:39.07 the user change the default to TPIOE=Y.

At 10:49:23.23, 10:49:33.52, 10:50:23.57, 10:50:33.81 MFNetDisk displays
message MPC049I (option 1) indicate that the devices are set to in error
status.

At 10:50:23.65, 10:50:33.88 MF return the DS P results and the devices are
displayed in error. All the pending IO requests were purged. No pending IO
missing errors are sent to the console.

10:51:18.66 after fixing the TCP connection, the user executes the F MPC,
DEV=ONL. This command will clear the errors status for the devices in error.


10:51:40.09 Do not forget to return the TPIOE=N setting option as the
recommended value for MFNetDisk and the full integrity of the data in
MFNetDisk disks



10:44:19.62 STC04950 00000090 MPC033I RECONECT TO PC ,<WRITE ,000032>

10:44:25.48 JJ02A 00000290 DS P,320,2

10:44:25.61 00000290 IEF196I IOS071I 0300,00,*MASTER*, CLEAR SUBCHANNE

10:44:25.61 STC04950 00000090 MPC033I RECONECT TO PC ,<WRITE ,000032>

10:44:25.61 00000090 IOS071I 0300,00,*MASTER*, CLEAR SUBCHANNEL INTERR

10:44:31.63 00000290 IEF196I IOS071I 0300,00,*MASTER*, CLEAR SUBCHANNE

10:44:31.61 00000090 IOS071I 0300,00,*MASTER*, CLEAR SUBCHANNEL INTERR

10:44:31.68 00000290 IEF196I IOS071I 0320,00,*MASTER*, CLEAR SUBCHANNE

10:44:31.67 00000090 IOS071I 0320,00,*MASTER*, CLEAR SUBCHANNEL INTERR

10:44:33.60 JJ02A 00000290 V 300,ONLINE

10:44:37.65 00000290 IEF196I IOS071I 0300,00,*MASTER*, CLEAR SUBCHANNE

10:44:37.63 00000090 IOS071I 0300,00,*MASTER*, CLEAR SUBCHANNEL INTERR

This is only part of the data. All data can be seen in MPCDOC.PDF using my
MFNetDisk code.

Thanks,
Shai

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html

Reply via email to