lvm cache volumes / flushing blocks

[rmalchow]

i have a cached lv and i want to extend it. afaik, i will have to remove the caching, extend the lv, and then add the caching again.

however:

lvconvert --uncache /dev/vgraid/lvraid

result in endless "flushing NNN blocks" and never finishes - same for other commands (such as lvremove vgraid/cache_data) . using lvs, i can see the following:

   LV                        VG     Attr       LSize    Pool                Origin         Data%  Meta%  Move Log Cpy%Sync Convert Devices                                                                    CacheTotalBlocks CacheUsedBlocks  CacheDirtyBlocks Type       CacheMode   
  [lvol0_pmspare]           vgraid ewi-------    2.00g                                                                            /dev/mapper/cache_data(359138)                                                                                                linear                 
  lvraid                    vgraid Cwi-aoC---   14.55t [lvraid_cache_data] [lvraid_corig] 99.99  0.58            99.32            lvraid_corig(0)                                                                      999340           999339           992494 cache      writethrough
  [lvraid_cache_data]       vgraid Cwi---C---    1.37t                                    99.99  0.58            99.32            lvraid_cache_data_cdata(0)                                                           999340           999339           992494 cache-pool writethrough
  [lvraid_cache_data_cdata] vgraid Cwi-ao----    1.37t                                                                            /dev/mapper/cache_data(0)                                                                                                     linear                 
  [lvraid_cache_data_cmeta] vgraid ewi-ao----    2.00g                                                                            /dev/mapper/cache_meta(0)                                                                                                     linear                 
  [lvraid_corig]            vgraid rwi-aoC---   14.55t                                                           100.00           lvraid_corig_rimage_0(0),lvraid_corig_rimage_1(0),lvraid_corig_rimage_2(0)                                                    raid5                  
  [lvraid_corig_rimage_0]   vgraid iwi-aor---   <7.28t                                                                            /dev/mapper/sda1_crypt(1)                                                                                                     linear                 
  [lvraid_corig_rimage_1]   vgraid iwi-aor---   <7.28t                                                                            /dev/mapper/sdb1_crypt(1)                                                                                                     linear                 
  [lvraid_corig_rimage_2]   vgraid iwi-aor---   <7.28t                                                                            /dev/mapper/sdc1_crypt(1)                                                                                                     linear                 
  [lvraid_corig_rmeta_0]    vgraid ewi-aor---    4.00m                                                                            /dev/mapper/sda1_crypt(0)                                                                                                     linear                 
  [lvraid_corig_rmeta_1]    vgraid ewi-aor---    4.00m                                                                            /dev/mapper/sdb1_crypt(0)                                                                                                     linear                 
  [lvraid_corig_rmeta_2]    vgraid ewi-aor---    4.00m                                                                            /dev/mapper/sdc1_crypt(0)                                                                                                     linear                 

one interesting thing i saw was that the cache_data lv is "not active":

  --- Logical volume ---
  Internal LV Name       lvraid_cache_data
  VG Name                vgraid
  LV UUID                3R19cf-WsX8-LJHA-4L9U-ytbw-x1Zo-YfU1Pl
  LV Write Access        read/write
  LV Creation host, time nasenhase, 2019-10-11 20:41:55 +0000
  LV Pool metadata       lvraid_cache_data_cmeta
  LV Pool data           lvraid_cache_data_cdata
  LV Status              NOT available
  LV Size                1.37 TiB
  Current LE             359138
  Segments               1
  Allocation             inherit
  Read ahead sectors     auto

how can i resovle this, short of copying everything off? when i do a vgcfgbackup, i get this:
vgraid_orig.txt

and if i modify it to this:
vgraid_modified.txt

the orig volume mounts fine, but it is missing data ... so apparently, something about the blocks it wants to flush is important after all.

what can i do to resolve this?

[zkabelac]

So few updates - recent version of lvm2 supports resize of cached LVs (>= 2.03.12)

If the uncaching does not proceed - it's likely a bug on 'origin volume' side - where there is most likely a write error so the cache cannot be decommissioned onto the origin storage and ATM it actually blocks further uncaching.

If this is your issue - please open a new issue/bug and provide also kernel trace (dmesg) of your system.