ombhurley@swift.cs.tcd.ie (10/17/90)
A friend of mine is working on a project and asked me to try and get some info for him. The project involves data aquisition, processing and storing and uses as its information core the Sybase product. It runs on a 25MHz 386 + 387 with 8M main mem and 110M hard storage. Sybase is installed on a raw partition that's about 30M in size. There is not extensive use made of interactive SQL, most of the database dealings involve use of library routines from continually and sporadically running C programs. (ie. some are continually logged into the database, others log in and out as necessary). p.s. the machine is running Interactive Unix 386 The problem(s) is(are) thus: Every now and then, and seemingly at random, a process that is logged in to the db will terminate correctly, but the db appears not to sever the login connection. The application code has been checked thoroughly and appears correct in view of the descriptions of the Sybase library functions. The effect of this is to cause the db, after an unspecified but short period of time to hang either for a matter of minutes or indefinitely. Sometimes there is an accompanying message saying that Sybase has run out of Pss structures. At this point the sql server is completely useless and must be brought down. If things hang completely, then this must be done violently, but most of the time isql and a bit of patience can allow the server to be shutdown correctly. Ok, server is down, so reboot server. This happens: --------------usual bumpf--------- kernel: network name /dev/tcp, port \x020007e97f0000001 kernel: ninit: transport provider: T_COTS kernel: ninit: t_bind, Couldn't allocate address kernel: ninit: t_unbind, 6, 0 server: Unable to initialize network 0 So the server doesn't boot. Netstat at this time says the following: (the name of the server is SQLSERVE) Active Internet connections Proto Recv-Q Send-Q Local Address Foreign Address (state) tcp 0 0 local.sqlserve local.1289 TIME_WAIT tcp 0 0 local.1287 *.* TIME_WAIT tcp 0 0 local.1285 *.* TIME_WAIT tcp 0 0 local.1281 *.* TIME_WAIT . . plus some more lines like the above. . Well then, can't start server so reset the machine. Server now runs properly (as you would guess). This problem is apparently a bit of a nuisance since the application is to run monitoring and logging 24 hours a day so the db+server gets a HUUUGE amount of use. Oh, BTW, I forgot to mention that the machine doesn't have an Ethernet card so it's using TCP/IP over the loopback driver. I wonder can anybody enlighten me to enlighten him. E-mail or news posting would be most welcome. Thankoo precognitively. O. -------------------------------------------------------------------------\ Oisin Hurley Dept. of C.S., T.C.D., Dublin 2, Ire. -----------------------> -------------- ombhurley@cs.tcd.ie --------------------------------------/
loyer@shalom.sybase.com (George Loyer) (10/23/90)
In article <7170.271c55b3@swift.cs.tcd.ie> ombhurley@swift.cs.tcd.ie writes: >A friend of mine is working on a project and asked me to try and get some info >for him. >The project involves data aquisition, processing and storing and uses as its >information core the Sybase product. It runs on a 25MHz 386 + 387 with 8M >main mem and 110M hard storage. Sybase is installed on a raw partition that's >about 30M in size. There is not extensive use made of interactive SQL, most >of the database dealings involve use of library routines from continually >and sporadically running C programs. (ie. some are continually logged into >the database, others log in and out as necessary). > >p.s. the machine is running Interactive Unix 386 > >The problem(s) is(are) thus: > ... This problem is complex enough that you should be talking to Sybase Technical Support (in London, for your Dublin friend). Here in the USA Technical Support Center we have seen at least one problem that has (somewhat) similar symptoms and for which a fix is available, but there are enough differences that your friend would benefit much from working with the Support Center. George Loyer Director, Technical Support Center Sybase, Inc. Emeryville, CA
lombardi@cshl.org (Diane Lombardi in Hershey) (03/07/91)
Hi Everyone, I apologize if this is the wrong newsgroup. I am using sybase DB-Libraries to update the tables in my database. I have imposed rules and triggers to support referential integrity. I want to send the updates as a batch process so that if one statement fails, then the entire batch is cancelled. This all works fine. What I need to know is what values are returned to dbsqlexec and dbresults when a rollback trans occurs in a trigger and when a rollback occurs in rule. I have no way of knowing if the process executed or was aborted. Your help would be greatly appreciated. Please email me at lombardi@cshl.org Thank-you Diane P.S. I am using sybase HyperCard DB-Library so I cannot simply put a message that prints to the screen inside the trigger, since the trigger is on a different platform.