Daniel Gryniewicz

Not really, no. If you have high enough logging on, you can examine the logs and figure it out, but we don't have auditing support in Ganesha.

Do you have a backtrace? Can you find, from the core, what was actually wrong? (null pointer, or whatever)

So you can't tell if it, for example, a NULL pointer deref, or a use-after-free, or anything like that? That makes it more difficult. Taking a guess as to the...

It was never committed, because it was never confirmed to fix the issue, and I couldn't reproduce locally. Submitted here: https://review.gerrithub.io/c/ffilz/nfs-ganesha/+/529440

This has come up in the past, with arguments on both sides. The argument that won was that adding a new export with a syntax error to a running Ganesha...

Those are not logged anywhere, and the krb5 error codes seem to be complicated enough that I can't figure out the message from the number by looking at code. It...

I was wrong, Matt pointed me to the list of well-know error codes here: https://andromeda.rutgers.edu/~sysmail/krb5_error.html This error is: -1765328203 | KRB5_KT_NOTFOUND | Key table entry not found This is presumably...

So, I'm not sure anyone has tested an HA setup with RGW. @mattbenjamin @jtlayton any thoughts?

HA should be fairly trivial for RGW, since you can't do arbitrary writes into files. Basically, set up HA like you would for CephFS, and it should work fine.

It looks like the FH cache in librgw isn't an LRU, but a tree that keeps every entry.