Thread View: vmsnet.sysmgt
2 messages
2 total messages
Started by Peter Smode
Fri, 22 Jan 1993 23:27
Lost LAT sessions
Author: Peter Smode
Date: Fri, 22 Jan 1993 23:27
Date: Fri, 22 Jan 1993 23:27
32 lines
1860 bytes
1860 bytes
We have a bit of an oddball problem happening at a couple of our sites. From time to time (measured in weeks or months), we will get one or more interactive terminal sessions on LTAxxxx: terminals getting "lost in space". The symptoms are as follows: - a SHOW USERS FULL will be missing the (server/port) address for the LTAxxx: terminal - the process will INHALE all available CPU time, at elivated priority. This effectively turns the system into a dog. - it is suspected that users terminal sessions just go dead in the water with no terminal response. We are not sure as users are not giving us good information when these situations occur. - SHOW ERROR reveals nothing - NCP counters do no show anything out of the ordinary - happens for both direct and dialup connections on any port, any server - server error counters do not seem to indicate trouble. more So far, DEC has sent the latest patch kits and had us install them, but to no avail. Has anybody got any ideas? Both system on which this occurs are homogeneous CI-based VAXClusters. CPUs include VAX 7xx, 8xxx and 6xxx based machines. VMS versions are 5.3-x and 5.4-x. -- Peter ************************************************************************** * Peter Smode E-mail: psmode@jeslacs.wimsey.bc.ca * * JES Library Automation Voice: (604)939-6775 * * Coquitlam, BC, CANADA Fax: (604)939-9970 * **************************************************************************
Re: Lost LAT sessions
Author: Dean Banfield
Date: Mon, 25 Jan 1993 22:20
Date: Mon, 25 Jan 1993 22:20
57 lines
3400 bytes
3400 bytes
In article <0NXsXB1w165w@jeslacs.wimsey.bc.ca> Peter Smode <psmode@jeslacs.wimsey.bc.ca> writes: > - a SHOW USERS FULL will be missing the (server/port) address for the > LTAxxx: terminal > - the process will INHALE all available CPU time, at elivated priority. > This effectively turns the system into a dog. > - it is suspected that users terminal sessions just go dead in the > water with no terminal response. We are not sure as users are not > giving us good information when these situations occur. We have a software product which experienced a similar problem when installed at one particular client. We narrowed it down to the following criteria: User logged on via LAT Hanging out in our product at our prompt (read wait on SYS$INPUT) User walks away, uses other session on terminal server, etc and forgets about his read prompt in our product. He then logs off terminal server without actually exiting from the product prompt, or logging off that session independently. LAT sends a HANGUP message to the process, which attempts to wind down the image. This is unsuccessful and the image is left disconnected from any LAT connection and gets very high priority in the attempted wind-down. We traced the problem to a set of CTL-C, CTL-Y trapping routines which we had implemented. In effect, the wind-down operation sent a CTL-C (or Y) to the image to wind it down. Our software, assuming that this was an interactive user, trapped the signal and tried to return the user to the product prompt. Having no effective terminal any more, the software got very confused and sucked down all available resources trying to re-issue the product prompt. To stop the bleeding (we never found an elegant solution) we simply advised the client to disable the CTL-C, CTL-Y trapping, which meant that when a user aborted out of our product, we allowed the system to wind it down in its own (possibly nasty to us) way. This DID stop the problem. My advice: Try to find out what application user/application was "turned off" in this way. Approach the developer of the application with these ideas and see what falls out. It may be difficult to track this down, but in our case our image was easily identifiable to the system manager as the one trying to wind itself down. The user was impossible to find, since the LAT stuff got disconnected early in the wind-down. Hope this is of some help. -Dean -- =============================================================== Dean Banfield Real Decisions Corporation Voice: 203.656.1500 22 Thorndal Circle FAX: 203.656.1659 Darien, CT USA 06820 Email: dean@rdcalr.realdec.com -- =============================================================== Dean Banfield Real Decisions Corporation Voice: 203.656.1500 22 Thorndal Circle FAX: 203.656.1659
Thread Navigation
This is a paginated view of messages in the thread with full content displayed inline.
Messages are displayed in chronological order, with the original post highlighted in green.
Use pagination controls to navigate through all messages in large threads.
Back to All Threads