| Title: | Discussion of CICS technical issues |
| Moderator: | IOSG::SMITHF |
| Created: | Mon Mar 13 1995 |
| Last Modified: | Fri Jun 06 1997 |
| Last Successful Update: | Fri Jun 06 1997 |
| Number of topics: | 192 |
| Total number of notes: | 680 |
Hi,
I have a question on control of connection from IBM clients.
We have about 1000 IBM clients and Dunix CICS.
If a region crashes and re-starts , then as soon as the region is up
at worst case,800 clients may try to connect to the region at the
same moment.
The region have 6 listeners(ld),therefore 6 ip port defined in the system.
So,if it is possile to start listeners(cicsip) manually then i can
start the 6 cicsip in a sufficient time interval so that can avoid the rushes
of client's requests.
Is it possible ?
p.s.: Of course,all requests will be queued within CICS , but i don't know
how many requests could be queued.
I have a experience that if the queued requests are much more than the
number of cicsas then the region fall into hang.
And we are impatient to wait the region to come back to normal state.
Thanks a lots,
dg.
| T.R | Title | User | Personal Name | Date | Lines |
|---|---|---|---|---|---|
| 185.1 | region startup delays... | LJSRV2::nqsrv422.nqo.dec.com::Workbench | Tue Apr 29 1997 15:11 | 25 | |
Hi, I don't know if this will help or not, but a customer I worked with had a similar situation with everyone trying to log back in simultaneously after a region crash, sometimes taking up to an hour to recover. In our case, the delay in the region starting was related to recovering all the aborted terminal sessions. If you are doing a warm re-start (and I presume you are...), you should see all the recovery procedures run for the database, the region and all previously logged in users, before the region is available for users to log in (all these events show up in the console.msg file...) We never did find a way to throttle the number of users logging in. We had ~150 users, and 20-30 cicsas servers running. Hint - if you haven't done this already, make sure you are starting all cicsas server processes up front when the region starts - don't allow it to happen dynamically, or you will likely see worse performance due to server startup overhead. I have not seen the problem you aluded to regarding the number of pending login requests queued .vs. cicsas servers running. rgds, Duane | |||||
| 185.2 | Any | DEKVC::JUNMOKSONG | Fri May 16 1997 19:00 | 7 | |
Hi,
I am in USA now and we , CICS instructor and me , discussed with
your problem and he said to check RegionPool Size ,that is , default is
1k. As you know, we can check that by CSTD on cicsterm.
I hope this help you.
Regards,
Jun-Mok Song
| |||||