*** Bug 1286207 has been marked as a duplicate of this bug. ***
RCA: dht_statfs_cbk should handle ENOENT/ESTALE errors just like dht_open_cbk/dht_stat_cbk as the file might've been moved out of the cached-subvol the client thinks. regards, Raghavendra
This issue still exists with glusterfs version: 3.8.4-13.el7rhgs.x86_64. Steps: 1) Created a ganesha cluster and created a distributed-replicate volume. 2) Enabled nfs-ganesha on the volume with mdcache settings. 3) Mounted the volume on multiple clients. 4) Created files and folders. 5) From mount point, issued rm -rf * and started removing bricks. During remove-brick rebalance, statfs failures are seen in rebalance logs. ERROR SNIPPET in rebalance logs: ================================ [2017-02-01 06:45:27.197342] E [MSGID: 109023] [dht-rebalance.c:779:__dht_check_free_space] 0-distrep-dht: failed to get statfs of /2/linux-4.4.1/drivers/net/wireless/ath/ath5k/ani.h on distrep-replicate-2 (No such file or directory)
This is probably a valid scenario if the file is question was deleted just as the statfs was sent. Do you see the issue with the steps listed in the description?
This issue is not reproduced on glusterfs version 3.8.4-15.el7rhgs.x86_64. Followed the same steps as in the description on both FUSE and NFS but the error reported in this BZ is not seen.