This is handled by the filesystem usually (or not, depending on what filesystem you use). When you hit a bad block you should just replace the drive - in case of a spinning disk the damage is likely going to spread, in case of flash device this error should have been prevented by firmware in the first place so I’d replace the drive rightaway and look at the rest of the same drives very suspiciously... But it is important to distinguish between manufacturing error and a mechanical failure on spinners - manufacturing error can be remapped and the drive can then work for years. But if you have an older drive that shows a bad block, just replace it.
Filesystems handle this differently, some have a bad block maps and some don’t, usually the error can be fixed on the drive itself by rewriting the offending sector (so filesystems don’t need to blacklist the block anyway) and it should do this on read as well. If the read fails you should run badblocks on the drive a few times and start with a fresh filesystem if you intend to keep it in the system. Jan > On 03 Aug 2015, at 16:01, 이영준 <youngjoon....@navercorp.com> wrote: > > Dear, Ceph. > > > I'm wondering how does Ceph isolate bad blocks when EIO occurs. > > I saw source codes and I found the logic of deep scrub, chunky_scrub() in > PG.cc. > > And I understood the real recovery logic is in submit_push_data() in > ReplicatedBackend.cc. > > It pulled an object from another replica and submit it. > > But, I couldn't find the logic about isolating bad blocks for preventing > future errors. > > I think EIO error will occur again when Ceph accesses that block block again. > > Is there any isolation logic for bad blocks? > > Thank you. > > > > 이 메일은 나눔글꼴로 작성되었습니다. 설치하기 <http://hangeul.naver.com/> > > > 이영준 LEE YOUNGJOON > Naver Labs 서비스플랫폼G 분산시스템개발랩 / 연구원 > > 경기도 성남시 분당구 불정로 6 네이버 그린팩토리, 463-867 > Tel 031-784-1078 Mobile 010-9300-2382 Fax 031-784-1000 > Email youngjoon....@navercorp.com <mailto:youngjoon....@navercorp.com> > Messenger <mailto:> > > <http://static.nhncorp.com/static/site/sign/> > _______________________________________________ > ceph-users mailing list > ceph-users@lists.ceph.com > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
_______________________________________________ ceph-users mailing list ceph-users@lists.ceph.com http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
