x6 ~ # lvconvert -y --type cache-pool --poolmetadata vga/cachemeta vga/cachedata WARNING: Unrecognised segment type cache-pool Invalid argument for --type: cache-pool Error during parsing of command line. After changing --with-cache=none to ...=internal in line 144 of ebuild: myconf="${myconf} --with-thin=none --with-cache=internal" # lvconvert -y --type cache-pool --poolmetadata vga/cachemeta vga/cachedata [...] Converted vga/cachedata to cache pool. Enabling +thin useflag would pull dev-libs/boost and it's an ostentation ;-) Conclusion: execute sed -i 's/\(with-cache=\)none/\1internal/' on sys-fs/lvm ebuilds Package grows 21KB/4M on amd64, setting with-cache=shared does not do the job.
If you simply change cache=none -> cache=internal, but do NOT have the cache_* binaries from thin-provisioning-tools, you're risking having problems in future. agk: as upstream, would you consider having cache=internal, but NOT the thin tools present to be a supported configuration?
> agk: as upstream, would you consider having cache=internal, but NOT the thin > tools present to be a supported configuration? Of course! I only don't want dev-libs/boost to take 110MB disk space (and 0.5h compilation...) on "thin" rescue system but still have cache support working. And it works like magic :-)
(In reply to Robin Johnson from comment #1) > If you simply change cache=none -> cache=internal, but do NOT have the > cache_* binaries from thin-provisioning-tools, you're risking having > problems in future. > > agk: as upstream, would you consider having cache=internal, but NOT the thin > tools present to be a supported configuration? Robin: any feedback here?
(In reply to Doug Goldstein from comment #3) > (In reply to Robin Johnson from comment #1) > > If you simply change cache=none -> cache=internal, but do NOT have the > > cache_* binaries from thin-provisioning-tools, you're risking having > > problems in future. > > > > agk: as upstream, would you consider having cache=internal, but NOT the thin > > tools present to be a supported configuration? > > Robin: any feedback here? I didn't hear anything back from agk at all.
I just ran into this issue, still seems to be showing up in sys-fs/lvm2-2.02.187-r2 . Is the current expectation to USE=thin and will I need to rebuild my custom initramfs with lvm2[thin] too, or will it be able to assemble it without [thin]?
while we wait for the maintainer to get back to us (5+ years), can someone please clarify: every use flag has to be cleared with an upstream maintainer as to whether it is a supported configuration? also in the meantime, would someone please add a USE flag for this configuration anyway, maybe masking it by default?
Ping, I applied said patch and am currently using a cache device without apparent issues. Had to recompile my kernel with CONFIG_DM_CACHE=y, though.
With both cache and thin targets, the userspace thin provisioning tools (which use boost) are used to make some basic checks on the device content before attempting to activate devices that use the targets - think of these tools as similar to 'fsck' for a filesystem - any repairing gets done in userspace. configure --with-cache-check etc. should control this for cache
That would mean that these checks are disabled, if I can use a cache with USE=-thin (lvdisplay contains a lot of cache data), no?