Daniel Gryniewicz

Hmm... I believe that leaves out latency in Ganesha itself, and means the latency must be in the kernel after we've written to the socket. @ffilz any ideas?

Please submit on gerrithub as described here: https://github.com/nfs-ganesha/nfs-ganesha/wiki/DevPolicy#Patch_Submission_Guidelines

The problem is that NFS ACLs are not Linux ACLs, and mapping between them is difficult. There is experimental code in FSAL_VFS to do this mapping, but it's not enabled...

It means an underlying filesystem returned EIO. This error (along with several others) translates to NFS4ERR_IO error, and this message is indicating that this translation has happened, and what the...

The first issue is the version. 2.7.3 is very old. At least upgrade to 2.7.6, if not 2.8.4, since there are many bug fixes in newer versions. Next, speculation. RGW...

Like I said, 2.7.3 is very old. Try something newer, many bugs have been fixed.

I'm not sure what python is doing; if the writes don't show up sequential at the server, they can't be treated as part of a single write, and so they...

There's tentative work to extend librgw to allow more complicated read/write patterns planned, but it hasn't been started yet. Since librgw is being used to implement a FUSE client, it...

Sorry, I forgot to add that FSAL_RGW was originally written to provide backup/restore and migration for RGW/S3, so the limitations weren't a problem.

Just to be clear, this isn't a limitation of Ganesha, it's a limitation of Ceph RGW. Ganesha provides full functional access to Ceph RGW, but RGW is not a full...