Daniel Gryniewicz

I thought we already had...

Those don't appear to be destroyed, it's true. They probably should be, but I don't think it ultimately makes much of a difference, since the memory is freed right after...

This sounds like a GFAPI bug. I would expect the client to either stay connected or reconnect in these circumstances. @kalebskeithley thoughts?

Unfortunately, Ganesha never removes a name from it's name-to-IP mapping.

Which other ops likely to block for significant amounts of time? In other words, is it worth doing async for other ops? We only did read and write because they...

I would hope that 1k entries in a readdir doesn't take long enough for async to be useful. If it does, I think we have bigger issues.

Readdir in ganesha doesn't read the entire directory in one go, it reads in chunks of size Dir_Chunk (default 128) with readahead if the FSAL supports it. Most FSALs that...

I'm not an expert on PROXY, but here's what seemed to happen. Ganesha tried to renew the session with the node 1, and failed with NFS4ERR_BADSLOT. PROXY only uses 16...

Things like this have usually been network issues in my experience. There are lots of potential network problems with IP failover. First, and most common, is the ARP table on...

The lease lifetime is related to the maximum latency that you may expect in your setup. That is, what is the longest that a client or server can ever take...