On 03/02/2016 01:40 PM, Philip S. Hempel wrote:
Hello everyone,
I am trying to repair a cluster that has 74 pgs that are down, I have
seen that the pgs in question are presently with 0 data on the OSD.
I have exported data from OSD's that were pulled when the client had
thought the disk were bad.
I am using the recovery method describe in "Incomplete PGs - OH MY!"
Following up with what also was stated in ceph users that you should put
the OSD out as well and set the weight correctly for the disk instead of 0.
I have done this running ceph 0.94.5 (the original ceph cluster was at
.80) and this is running on a Proxmox server at version 3.4.
I have imported some of the data into a temp OSD, the PG's will import,
but some of them cause the OSD to segfault like this below.
ceph version 0.94.5 (9764da52395923e0b32908d83a9f7304401fee43)
1: /usr/bin/ceph-osd() [0xbf03dc]
2: (()+0xf0a0) [0x7fe288b640a0]
3: (gsignal()+0x35) [0x7fe2874cc125]
4: (abort()+0x180) [0x7fe2874cf3a0]
5: (__gnu_cxx::__verbose_terminate_handler()+0x11d) [0x7fe287d2489d]
6: (()+0x63996) [0x7fe287d22996]
7: (()+0x639c3) [0x7fe287d229c3]
8: (()+0x63bee) [0x7fe287d22bee]
9: (ceph::__ceph_assert_fail(char const*, char const*, int, char
const*)+0x220) [0xcddda0]
10: /usr/bin/ceph-osd() [0x7f578f]
11: (pg_interval_t::check_new_interval(int, int, std::vector<int,
std::allocator<int> > const&, std::vector<int, std::allocator<int> >
const&, int, int, std::vector
<int, std::allocator<int> > const&, std::vector<int, std::allocator<int>
> const&, unsigned int, unsigned int, std::tr1::shared_ptr<OSDMap
const>, std::tr1::shared_p
tr<OSDMap const>, pg_t, IsPGRecoverablePredicate*, std::map<unsigned
int, pg_interval_t, std::less<unsigned int>,
std::allocator<std::pair<unsigned int const, pg_int
erval_t> > >*, std::ostream*)+0x2ba) [0x8c399a]
12: (OSD::build_past_intervals_parallel()+0xbe1) [0x7d2261]
13: (OSD::load_pgs()+0x2d8a) [0x7e977a]
14: (OSD::init()+0xdac) [0x7ebb2c]
15: (main()+0x253e) [0x78dd6e]
16: (__libc_start_main()+0xfd) [0x7fe2874b8ead]
17: /usr/bin/ceph-osd() [0x793de9]
NOTE: a copy of the executable, or `objdump -rdS <executable>` is
needed to interpret this.
Now none of the data imported is considered by Ceph as possible data for
the PG's that are incomplete, so this is one of the main problems I am
trying to rectify.
I have set after the upgrade of Ceph the tunables to optimal
I have tried repairs on all incomplete PG's.
I have tried scrub and deep-scrub on incomplete PG's.
What I would hope to accomplish is the PG data I do have, I could get
complete PG's from.
Thanks
Please let me know what other data I could give to help determine a fix
for this.
I am looking to hire a consultant to support us with this.
Please respond so we can get an agreement made and I can pass more
details on to you.
Thanks again.
_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
