Operating System - OpenVMS
1748073 Members
5364 Online
108758 Solutions
New Discussion

Re: Availability Manager loose paths after 5-10 min

 
SOLVED
Go to solution
RainerWagner
Occasional Advisor

Availability Manager loose paths after 5-10 min

Hi @ll,

 

our Data Analyzer V3.1-2 on Windows XP lost OpenVMS paths after 5-10 min runtime.

We measure five I64 systems and six Alpha systems. These system are all in default LAN and will be measured by one Data Server in the same LAN. The Data Analyzer are opened in other LAN's.

The functionality is there, but terminates after a few minutes.

Data Analyzer manually startet with system account $ avail/server . Parameter /iheap and /mheap had not been successfully, so we uses default values.

After temination it is not possible to restart data analyzer (you see no nodes), restart data server is also not successful. We have to stop amds and start amds on data server node and then it is running again (for about 10min).

Are there some parameter like TCPIP or account parameter to solve this problem ?

These parameters we have as system:

Maxjobs:         0  Fillm:      4096  Bytlm:      4000000
Maxacctjobs:     0  Shrfillm:      0  Pbytlm:           0
Maxdetach:       0  BIOlm:       150  JTquota:       4096
Prclm:          10  DIOlm:       150  WSdef:         4096
Prio:            4  ASTlm:       300  WSquo:         8192
Queprio:         0  TQElm:       100  WSextent:     16384
CPU:        (none)  Enqlm:     32767  Pgflquo:  280000000

kind regards

Rainer Wagner

11 REPLIES 11
Volker Halle
Honored Contributor

Re: Availability Manager loose paths after 5-10 min

Rainer,

 

did this ever work ? What has changed ?

 

From reading your problem description, it looks like a problem on the data server node, either in the Availabilty Manager data server or in the AMDS driver on that node.

 

Does the data server have any logfile ? Any messages in there ? Anything in OPCOM on the data server ?

 

Does the AMDS driver continue to send and receive AMDS packets over the ethernet ? You can check with SDA:

 

$ ANALYZE/SYSTEM

SDA> SHOW LAN                         ! look for the UCB name with Client = AMDS

SDA> SHOW LAN/DEV=<amds-ucb-name>

 

Look for the counter values.

 

Volker.

RainerWagner
Occasional Advisor

Re: Availability Manager loose paths after 5-10 min

Volker,

 

thanks for your helping.

I watched the data server and I give you some more information in attached logfiles. Sent counters increased after stopping analyzer.

 

Can you look into these values ?

 

kr Rainer

Duncan Morris
Honored Contributor

Re: Availability Manager loose paths after 5-10 min

The SDA counters are showing:

 

Unavail user buffer              218

 

This looks like the problem I reported to HP concerning changes that were implemented in the LAN drivers (from VMS732_LAN-V0700 onwards). Unfortunately, I cannot find the original thread in this new itrc!

 

The necessary changes to SYSGEN parameters to get round the problem are now documented in the VMS 8.4 release notes, section 4.5:

 

http://h71000.www7.hp.com/doc/84final/6677/6677pro_sm.html#avail_man_31

 

  • The Availability Manager Analyzer reports "Path Lost" PATHLST events for all the remote nodes and stops displaying data after some elapsed time.
    Workaround:
    The workaround for this problem is to modify LAN_FLAGS bit to 16 in SYSGEN parameter, which restores normal behavior. The command is as follows:
    $ MC SYSGEN SET LAN_FLAGS 16 

 

 

 

Volker Halle
Honored Contributor

Re: Availability Manager loose paths after 5-10 min

Duncan,

 

this note from the V8.4 release notes looks VERY questionable !

 

Setting LAN_FLAGS to 16 will set bit 4 in the LAN_FLAGS system parameter longword. According to SYSGEN HELP SYS_P LAN_FLAGS, bit 4 corresponds to setting some attributes on ATM adapters !

 

Maybe the note should read: ... modify LAN_FLAGS bit 16 in SYSGEN parameter. Especially as the release notes of VMS732_LAN-V0700, you were referring to, talks about bits 14 and 15 in the context of the new 'LAN Event Reporting' feature.

 

One could try setting bit 16 in LAN_FLAGS with SYSGEN> SET LAN_FLAGS %X00010000 (followed by a WRITE ACTIVE) to see, if that would solve the problem.

 

Hoffmanlabs also has some info about this, which seems to support my concerns:

 

http://labs.hoffmanlabs.com/node/1709

 

Volker.

Volker Halle
Honored Contributor

Re: Availability Manager loose paths after 5-10 min

Rainer,

 

from looking at your SDA_SHOW_LAN.TXT output, I would conclude, that the SYS$RMDRIVER is not receiving AMDS packets nor AMDS multicast PDUs since 11:27:30 on LAN interface EIA0 !

 

EIA27:     Last receive         21-JUL 11:27:30     <<< no more AMDS packets received after that time

 

These values remained the same after 11:34 (supporting observation above):

 

PDUs received                       5277   

Mcast PDUs received              352

 

Those values increased after 11:34:

 

Unavail user buffer                   218

Last UUB time        21-JUL 11:34:27   

 

Volker.

 

 

RainerWagner
Occasional Advisor

Re: Availability Manager loose paths after 5-10 min

Duncan, Volker,

 

thanks for your replies.

With LAN_FLAGS 16 the problem occured again for all nodes except the node where data analyzer runs and I updated LAN_FLAGS.

Now I am testing the hex parameter.

$> mc sysgen
SYSGEN>  SHOW LAN_FLAGS
Parameter Name            Current    Default     Min.       Max.   Unit  Dynamic
--------------            -------    -------   -------    -------  ----  -------
LAN_FLAGS                      16          0         0         -1 Bitmask    D
SYSGEN>  USE CURRENT
SYSGEN>  SHOW .    
Parameter Name            Current    Default     Min.       Max.   Unit  Dynamic
--------------            -------    -------   -------    -------  ----  -------
LAN_FLAGS                      16          0         0         -1 Bitmask    D
SYSGEN>  SET LAN_FLAGS %X00010000
SYSGEN>  WRITE ACTIVE
SYSGEN>  SHOW .
Parameter Name            Current    Default     Min.       Max.   Unit  Dynamic
--------------            -------    -------   -------    -------  ----  -------
LAN_FLAGS                   65536          0         0         -1 Bitmask    D
SYSGEN>  EXIT

Rainer

Duncan Morris
Honored Contributor

Re: Availability Manager loose paths after 5-10 min

Volker,

 

you are quite right.

 

That was why I was trying to locate my original thread - I had a niggling doubt about that flag setting!

I am away from the systems where I had implemented the LAN_FLAGS changes, so I could not check the MODPARAMS data, but I did seem to remember setting a large mask value rather than 16.

 

RainerWagner
Occasional Advisor

Re: Availability Manager loose paths after 5-10 min

it looks good. :smileyvery-happy:

RainerWagner
Occasional Advisor

Re: Availability Manager loose paths after 5-10 min

Volker, Duncan,

 

I think upgrade LAN_FLAGS is required only on node with running Data Server.

Now Analyzer is running well (since 20min) and I 'll close my question as solved.

 

Thank you very much for your solutions. I am be glad.

 

Rainer