[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: SLAP_LIGHTWEIGHT_LISTENER, using lazy_sem is a bad design

The title sounds like a tabloid headline... :-)

Howard Chu wrote:

This solution will deadlock if all worker threads are stuck in a write wait. Since the semaphore completely blocks the listener thread, there will be no way to wake up the waiting writers and free up more threads. Also, blocking the listener thread like this prevents the idletimeout checker from working. I.e., you have managed to disable two key mechanisms for returning server resources to the pool, precisely when they are needed the most.

The listener thread must never block, period.

It would be better to simply have ldap_pvt_thread_pool_submit return a result code (e.g. LDAP_BUSY if the submitted op will be queued because there are no available workers, LDAP_SUCCESS otherwise) that is passed back to the listener thread. When the listener thread gets this result it should drop all read descriptors from the event set, but keep monitoring the wake_sds and the write events.

Alternatively we should stop monitoring write events and just let the writers unblock themselves. Is there a particular reason why we monitor write events? I don't see any benefit. Eliminating one set of event sources would reduce our kernel load by half.

I want to ask the same question, too. What was the original rationale behind the listener's waking up the write-wait workers via condition variable ? If there're no compeling reason to have the listener wake up the writers (this certainly does nothing to do with concurrency improvement), we should consider the idea of relying on the writers themselves and the OS to unblock.

Aside from this issue there is definitely a bug in the current implementation; I see the same event being submitted multiple times in rapid succession. The CPU usage goes to 100% and there does not appear to be any end condition that disables the event. This occurs most often in test033, but should occur in any test that uses syncrepl (or listener-managed client tasks like syncrepl). After the syncrepl task has sent a search request to the provider and the first reply arrives, marking the socket readable. It appears this readable state is not getting reset.

Will look into this further.