Reproducibility :- Easy A replicated volume of 2 nodes and 2 backends with replica count 2. Is expanded to distributed replicated volume of 3 x 2. After the expansion the VM's were running during the expansion and there was seamless reflection of additional space dynamically without a remount. Started the rebalance and after like 30 files transferred stopped in bet'n to see if the files are reflected properly. Started seeing stale file handle errors for certain directories which were still being accessed by the VM's. After a remount the issue goes away. Why would a user see this error? does rebalance change the gfids for files? Does directory layout change not properly reflected on to clients?
(In reply to comment #0) > Reproducibility :- Easy > > A replicated volume of 2 nodes and 2 backends with replica count 2. Is expanded > to distributed replicated volume of 3 x 2. > > After the expansion the VM's were running during the expansion and there was > seamless reflection of additional space dynamically without a remount. Will this be fixed in 3.1.2 release? This is tracked for a customer issue.
(In reply to comment #0) > Reproducibility :- Easy > Reproducibility was not at all easy for us. Instead, we couldn't even reproduce it (with latest git (say from last 7 days)). Will be closing it with 'works for me'. Please reopen if it exist in codebase after two pending rebalance patches get in.