Boost.Log and reinitializing text file backend on I/O errors
I'm using text file sink to log into a file on NFS filesystem. I would like to make logging more robust when it comes to unreliability of NFS.
I've registered an exception handler to ignore I/O errors, but I don't know how to re-open the log file automatically so that logging would continue when the filesystem is online again. My first try was to remove and reinitialize all sinks when the exception handler is called, but this seems to cause deadlock.
Is there a way to simply re-open the associated file which caused the exception?