Wt

The beginning of the discussion is here

The Wt fcgi Server implementation has this code:

void Server::handleSigChld()
{
    ...
    // shared process option

    static int childrenDied = 0;

    ++childrenDied;

    if (childrenDied < 5)
        spawnSharedProcess();
    else
        LOG_ERROR_S(&wt_, "sessions process restart limit (5) reached");

    ...
}

The number 5 should be configurable at least.

But I think that it will be much better to have some extended crash counter. 5 crashes within 5 days is not the same as 5 crashes within 5 seconds. So I suggest to track the number of crashes per some interval. In this case if we exceed some configurable value (i.e. 10 crashes within 10 last seconds) we should start spawning processes with some delay (5-10 seconds). It will be better than to completely stop spawning new processes.