[Linux-cluster] GFS inodes

[Steffan Henke]

On Fri, 20 Jul 2007, David Teigland wrote:

> This is strange, after being unlinked, gfs_inoded should deallocate them,
> but apparently it's not.  What release are you using?  Could you record

Hello,

I am using EL 4.5 with kernel 2.6.9-55.ELsmp.

> the inode numbers of the files before they're unlinked next time (ls -li)?

This is my current dir prior to the mv:

total 2222900
  590019 -rw-r--r--  1 2271684518 Jul 23 06:57 _2cqqt.cfs
4587612 -rw-r--r--  1          4 Jul 23 06:57 deletable
4587613 -rw-r--r--  1         31 Jul 23 06:57 segments

> Then, after they're unlinked, do 'gfs_tool counters' (looking for unlinked
> count), and 'gfs_tool lockdump' and send this info?  I'd like to see what

gfs_tool counters /gfs

                                   locks 426
                              locks held 32
                            freeze count 0
                           incore inodes 10
                        metadata buffers 221
                         unlinked inodes 129
                               quota IDs 2
                      incore log buffers 0
                          log space used 0.05%
               meta header cache entries 0
                      glock dependencies 0
                  glocks on reclaim list 0
                               log wraps 1
                    outstanding LM calls 0
                   outstanding BIO calls 0
                        fh2dentry misses 0
                        glocks reclaimed 1914903
                          glock nq calls 3738433
                          glock dq calls 3641879
                    glock prefetch calls 1120342
                           lm_lock calls 1236087
                         lm_unlock calls 1139499
                            lm callbacks 2375643
                      address operations 1700121
                       dentry operations 44614
                       export operations 0
                         file operations 22763
                        inode operations 2365329
                        super operations 2265685
                           vm operations 0
                         block I/O reads 582429
                        block I/O writes 576463

gfs_tool lockdump returns an error message on all cluster nodes:
gfs_tool: unknown mountpoint /gfs

> state the iopen locks are in for the unlinked files.  Also send the output
> of 'gfs_tool gettune'.

gfs_tool gettune /gfs
ilimit1 = 10
ilimit1_tries = 3
ilimit1_min = 10
ilimit2 = 10
ilimit2_tries = 10
ilimit2_min = 3
demote_secs = 300
incore_log_blocks = 1024
jindex_refresh_secs = 60
depend_secs = 60
scand_secs = 5
recoverd_secs = 60
logd_secs = 1
quotad_secs = 5
inoded_secs = 15
glock_purge = 0
quota_simul_sync = 64
quota_warn_period = 10
atime_quantum = 3600
quota_quantum = 60
quota_scale = 1.0000   (1, 1)
quota_enforce = 1
quota_account = 1
new_files_jdata = 0
new_files_directio = 0
max_atomic_write = 4194304
max_readahead = 262144
lockdump_size = 131072
stall_secs = 600
complain_secs = 10
reclaim_limit = 50
entries_per_readdir = 32
prefetch_secs = 10
statfs_slots = 64
max_mhc = 10000
greedy_default = 100
greedy_quantum = 25
greedy_max = 250
rgrp_try_threshold = 100
statfs_fast = 0

A gfs_tool reclaim returns *some* metadata, but no inode at all, while a
service gfs restart shows

Jul 23 09:54:59 host kernel: GFS: fsid=gfs:lucene.0: Scanning for 
log elements...
Jul 23 09:54:59 host kernel: GFS: fsid=gfs:lucene.0: Found 130 
unlinked inodes

- however, they are not de-allocated unless I restart GFS on some/all 
cluster nodes...

Regards,

Steffan