Hi Kern,
Thanks for the reply. The " Specified slot ignored." error was my fault - I tried to mount to a drive which was already loaded. I did run through those steps again with success, but copied the wrong 'mount; results into my email.
I labeled all the tapes with their virtual bar codes using 'label barcodes' then 'update slots'. The label listed using mtx for this volume is correct and it was working well for several months so I assume it has since been corrupted.
Your comment on connectivity in the could resonates with me based on my experience using Bacula with AWS storage gateway for 6-9 months. If the gateway is running 24/7 then I have zero connectivity issues and zero device errors. This is one reason I decided to locate the Bacula server on the same subnet in AWS as the storage gateway - the connection is extremely responsive. I think that having the Bacula server in our local network and communicating to a remote AWS storage gateway would present more connectivity problems.
Your comment about Bacula assuming that all I/O works really seems to me to be the source of my current problem. Since I started shutting the storage gateway down every day from 2:30 AM - 10:30 PM I've noticed that sometimes the iscsi devices do no reconnect from the Bacula server to the Gateway. Or, if they do reconnect they sometimes have conflicting or incorrect iscsi device mapping. So far, I have noticed that this kind of error effects the mapping of the drive changer being listed on the same path as a drive. A reboot of the Bacula server usually clears this up.
However, if there is some kind of iscsi mapping problem and Bacula starts i'ts jobs I can see how this would result in all kinds of problems. I'm assuming this is what happened here.
I could try to write a shell script on the Bacula server which checks iscsi mapping before Bacula starts it's nightly jobs. Then, if the shell script notices a device mapping problem and can't fix it, it would shut down Bacula and email me. This wouldn't solve the mapping problem, but it might help me avoid corrupted volumes because I could login and correct mapping and then kick the Bacula jobs off. What do you think?
Thank you!