Ok.. I solved..
I just detached and reattached the cache

# echo 7eb257b3-940d-42ca-ab23-52752f8b17f8 > /sys/block/bcache1/bcache/detach
# echo 7eb257b3-940d-42ca-ab23-52752f8b17f8 > /sys/block/bcache1/bcache/attach

that did the trick.

# bcache-super-show /dev/sdd1
sb.magic		ok
sb.first_sector		8 [match]
sb.csum			98719AA13E8D185F [match]
sb.version		1 [backing device]
dev.label		(empty)
dev.uuid		904aaaa4-473a-446d-aad5-4e55cde972a8
dev.sectors_per_block	1
dev.sectors_per_bucket	1024
dev.data.first_sector	16
dev.data.cache_mode	0 [writethrough]
dev.data.cache_state	1 [clean]
cset.uuid		7eb257b3-940d-42ca-ab23-52752f8b17f8

# bcache-super-show /dev/sde1
sb.magic		ok
sb.first_sector		8 [match]
sb.csum			98719AA13E8D185F [match]
sb.version		1 [backing device]
dev.label		(empty)
dev.uuid		904aaaa4-473a-446d-aad5-4e55cde972a8
dev.sectors_per_block	1
dev.sectors_per_bucket	1024
dev.data.first_sector	16
dev.data.cache_mode	0 [writethrough]
dev.data.cache_state	1 [clean]
cset.uuid		7eb257b3-940d-42ca-ab23-52752f8b17f8


Everything is ok.

Il Fri, 07 Oct 2016 12:55:50 +0000, Michele Brodoloni ha scritto:

Hi,
I tried to reboot the machine, but bcache is still dead.
/sys/block/bcache0/bcache/state reports "clean" (RAID5)
/sys/block/bcache1/bcache/state reports "no cache" (SAN RAID10)

I did not mention before that the SAN is used like a DAS, and it isn't
shared with other machines. There's just one server accessing it.

Regards,
Michele

Il Fri, 07 Oct 2016 10:46:39 +0000, Michele Brodoloni ha scritto:

quoted

Hello,
I have bcache running on a Debian 8.0 x86_64 with kernel 4.4.16.
I got 2x Samsung PRO 850 250gb in hardware RAID0 acting as cache device
for 1) a local RAID5 volume 2) a SAN RAID10 volume in active/backup
multipath fiber channel

I noticed today that SAN backing devices got detached from cache:

# bcache-super-show /dev/sdc1 sb.magic		ok

sb.first_sector		8 [match]

quoted

sb.csum			2D71F678442855F6 [match]
sb.version		3 [cache device]
dev.label		(empty)
dev.uuid		c3dd7b4e-04e0-4578-a0ce-b35a5745e459

dev.sectors_per_block	1

quoted

dev.sectors_per_bucket	1024 dev.cache.first_sector	1024
dev.cache.cache_sectors	629144576 dev.cache.total_sectors

629145600

quoted

dev.cache.ordered	yes dev.cache.discard	no

dev.cache.pos		0

quoted

dev.cache.replacement	0 [lru] cset.uuid
7eb257b3-940d-42ca-ab23-52752f8b17f8

# bcache-super-show /dev/sdd1 sb.magic		ok

sb.first_sector		8 [match]

quoted

sb.csum			514C0F59BC7C1938 [match]
sb.version		1 [backing device]
dev.label		(empty)
dev.uuid		904aaaa4-473a-446d-aad5-4e55cde972a8

dev.sectors_per_block	1

quoted

dev.sectors_per_bucket	1024 dev.data.first_sector	16

dev.data.cache_mode

quoted

0 [writethrough]
dev.data.cache_state	0 [detached] cset.uuid
00000000-0000-0000-0000-000000000000

# bcache-super-show /dev/sde1 sb.magic		ok

sb.first_sector		8 [match]

quoted

sb.csum			514C0F59BC7C1938 [match]
sb.version		1 [backing device]
dev.label		(empty)
dev.uuid		904aaaa4-473a-446d-aad5-4e55cde972a8

dev.sectors_per_block	1

quoted

dev.sectors_per_bucket	1024 dev.data.first_sector	16

dev.data.cache_mode

quoted

0 [writethrough]
dev.data.cache_state	0 [detached] cset.uuid
00000000-0000-0000-0000-000000000000

(Obviously /dev/sdd1 and /dev/sde1 are the same device but with
different path)

Another side effect is that my syslog, kern.log logfiles are eating all
my root partition space with messages like this:

Oct  7 12:27:31 lnx kernel: [2300151.278097] bcache: bch_btree_gc() gc
failed!


What can be done to troubleshoot this issue?

Regards,
Michele