Greetings. We've recently been seeing several database crashes. The last things seen in the logs are : postmaster: StreamConnection: accept: Too many open files in system postmaster: StreamConnection: accept: Too many open files in system FATAL 1: ReleaseLruFile: No open files available to be closed IpcSemaphoreLock: semop failed (Identifier removed) id=512 IpcSemaphoreLock: semop failed (Identifier removed) id=513 (repeated about 20-30 times) IpcSemaphoreUnlock: semop failed (Invalid argument) id=515 FATAL: s_lock(40014065) at spin.c:127, stuck spinlock. Aborting. FATAL: s_lock(40014065) at spin.c:127, stuck spinlock. Aborting. FATAL: s_lock(40014065) at spin.c:127, stuck spinlock. Aborting. (repeated about 100 times) This is PostgreSQL v7.0.2 under a Debian 2.2 system with a custom 2.2.17 kernel installed. I have the maximum per-process file descriptor limit set to 1024, and the maximum system file descriptor limit set to 8192, so I'm confused about the first messages. I'm also seeing : Sorry, too many clients already Sorry, too many clients already Sorry, too many clients already notices, but I thought I increased the number of allowed conncetions. What's the correct way to go about doing this? :)
