Return-Path: pg_adm@postgres.berkeley.edu Received: by postgres.Berkeley.EDU (5.61/1.29) id AA20245; Thu, 18 Feb 93 18:39:31 -0800 Date: Thu, 18 Feb 93 18:39:31 -0800 Message-Id: <9302190239.AA20245@postgres.Berkeley.EDU> From: Carole Goble Subject: a problem using Postgres on Suns To: postgres@postgres.berkeley.edu Sender: pg_adm@postgres.berkeley.edu Cc: cag@computer-science.manchester.ac.uk Dear Postgres experts We have recently installed Postgres on our Sun cluster for a number of students to explore. We do, however, have a problem which we think is a mistake in our install and we would appreciate some advice. We are running the postmaster on a Sun Sparc 2 with 32 Mb memory. We have thus installed the TTMOREIPCS in order to run more Postgres backends. The clients are Sparcs. We get this situation: 10+ students are running backends on their own databases using monitor. They work for about an hour or so retrieving, defining functions etc.. Then they begin to fail on define statements, retrieves using user-defined functions, replace and append statements. Straightforward retrieves continue to work. The error is that the postmaster has signalled an error to the backend and terminated the backend unexpectedly. The user is returned to the unix prompt. The backend is still running on the server however. A warning that shared memory may be corrupted is also given. (sorry I haven't a copy of the exact wording here). The error is NOT the complaint about the lack of shared memory warned of in the manual. Once this happens they all seem to collapse. An ipcclean helps for a short while, as does rebooting the server. If, however, a few (say 4) students use Postgres simultaneously it stays up without a problem. Postgres is running on the same server as Oracle (no more space anywhere else!) but no one is using Oracle when the Postgres users are on. Thanking you in anticipation with kind regards Carole Goble University of Manchester UK