Regain control when #maxusers is reached

[oldemanw]

Hi All,

I'm faced here with a "frozen" database that I try to regain control on.
Progress 10.1c02, server started with:

$DLC/bin/proserve /mfg/data/db2/mfgprod -N TCP -H nlalm1mfg01 -S qaddb_prd -Mm 32600 -L 200000 -c 350 -B 80000 -bibufs 96 -n 240 -spin 16000 -Mf 60 -Mxs 32000 -aibufs 144 -cpinternal UTF-8 -cpcoll basic -Mn 240 -ServerType 4GL

It seems the max users (-n parameter) was hit, for users that now try to connect the db appears frozen.

Any suggestions on how to regain control? Promon freezes too.

Thanks in advance,
Willem.

 

[oldemanw]

Ah, you may want to see this:

[mfg@nlalm1mfg01 ~]$ proutil -C dbipcs
OpenEdge Release 10.1C02 as of Fri Sep 19 21:44:04 EDT 2008

PROGRESS SHARED MEMORY STATUS
ID ShMemVer Seg# InUse Database
0 - - - (not PROGRESS)
32769 - - - (not PROGRESS)
65538 - - - (not PROGRESS)
98307 10213 0 Yes /mfg/data/db2/mfgprod.db
131076 10213 0 Yes /mfg/data/db2/admprod.db
163845 10213 0 Yes /mfg/data/db2/audprod.db
196614 10213 0 Yes /mfg/data/brasil/brasildb.db
229383 10213 0 Yes /mfg/data/tms2/xttraxmu.db


[mfg@nlalm1mfg01 ~]$ ipcs -a

------ Shared Memory Segments --------
key shmid owner perms bytes nattch status
0x00000000 0 root 644 40 2
0x00000000 32769 root 644 16384 2
0x00000000 65538 root 644 268 2
0x00000000 98307 mfg 644 746766336 334 dest
0x00000000 131076 mfg 644 144654336 43
0x00000000 163845 mfg 644 12730368 39
0x00000000 196614 mfg 664 8445952 4
0x00000000 229383 mfg 644 101203968 41

------ Semaphore Arrays --------
key semid owner perms nsems
0x00000000 32769 mfg 666 7
0x00000000 65538 mfg 666 241
0x00000000 98307 mfg 666 241
0x00000000 131076 mfg 666 7
0x00000000 163845 mfg 666 123
0x00000000 196614 mfg 666 123
0x00000000 229383 mfg 666 7
0x00000000 262152 mfg 666 123
0x00000000 294921 mfg 666 123
0x00000000 327690 mfg 666 7
0x00000000 360459 mfg 666 123
0x00000000 393228 mfg 666 123
0x00000000 425997 mfg 666 7
0x00000000 458766 mfg 666 134
0x00000000 491535 mfg 666 134

------ Message Queues --------
key msqid owner perms used-bytes messages

 

[TomBascom]

You need to kill some users.

 

[TomBascom]

Or get them to log out peacefullly.

And then you need to increase -n.

Always set -n well in excess of your actual user count. It does not equal licensed users. -n has nothing to do with licenses.

 

[oldemanw]

Hi Tom,

That's my problem, killing users (or nicely asking :biggrin doesn't help. For one reason or another Progress still thinks they are connected.
In my .lg file I find lots of "HANGUP singnal received messages" and a "who -a" displays many lost sessions.


Still investigating, thanks for reading though.

Willem.

 

[TomBascom]

HANGUP "swarms" are usually the result of a network error.

You may just need to force a shutdown and restart.

 

[TomBascom]

FWIW you have -n = -Mn and you do not have -Ma set which means that it is defaulted to 5. -Mi is defaulting to 1. In combination this means that every remote user is spawning a server until you run out of connections.

A more typical set of parameters for a 240 licensed user system would be:

-n 500
-Mn 50
-Ma 5
-Mi 1

 

[oldemanw]

Tom,

Thanks a lot for that hint about the network.
Our admin has been playing with DNS (had to sit on him to get a confession) and indeed the log is swarmed with HANGUPS...
Will keep you updated on the outcome. I'll be back tomorrow.

Willem

 

[oldemanw]

I set Mn = 40 and added Ma and Mi. We peak around 115 users every now and then, so n = 240 should be plenty for us.
That suggestion was also welcome It will be in effect next week after restarting the server.

The machine is happily churning away at the moment.