Todd Lipcon has posted comments on this change. ( 
http://gerrit.cloudera.org:8080/8048 )

Change subject: KUDU-2135 (part 1): add persistent disk states
......................................................................


Patch Set 7:

(2 comments)

Should this be abandoned? I think it was re-posted elsewhere or made obsolete 
by design changes?

(apparently I also had some draft comments on this which are going to be posted 
with the above question, feel free to ignore them)

http://gerrit.cloudera.org:8080/#/c/8048/7//COMMIT_MSG
Commit Message:

http://gerrit.cloudera.org:8080/#/c/8048/7//COMMIT_MSG@13
PS7, Line 13:  failed disks will not
            : be used and that data is where it is expected be (which will be
            : increasingly important as we begin to supporting addition, and
            : replacement of disks).
Can you expand on this a bit in the commit message? I'm not entirely sure I 
understand what "data is where it is expected to be" means. For example, 
imagine I have disks mounted at /data/{1,2,3}/ and on reboot for some reason 
the disks are re-assigned such that the mount point /data/1 refers to the data 
that used to be on /data/3/. Is that prevented by this?

FWIW the above isn't that out-of-the-ordinary -- the assignment of 
/dev/sd{a,b,c} at boot time can be non-deterministic in some setups (or at 
least it was when I last did some ops stuff). Best practice is to use "UUID=" 
mappings in /etc/fstab so that the disks show up in consistent spots but I'm 
not sure everyone does this, so would be a good idea to know ahead of time what 
would happen.


Similarly, what if for some reason I decided that instead of mounting at 
/data/{1,2,3} I wanted to mount at /data/{1,2,3}/kudu or something (i.e change 
the paths but keep the same blocks in the same dirs). Would that be allowed or 
prohibited? Why?

Again the above isn't that crazy of a scenario, eg if someone starts with using 
a disk just for Kudu and then decides they also want to install HDFS and thus 
moves all the kudu data into a subdir.


http://gerrit.cloudera.org:8080/#/c/8048/7//COMMIT_MSG@19
PS7, Line 19: to never use a previously failed disk,
I think I've been confused about this several times but still can't quite 
remember the reasoning here. Why is it bad to re-attempt to open a 
previously-failed disk?



--
To view, visit http://gerrit.cloudera.org:8080/8048
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings

Gerrit-Project: kudu
Gerrit-Branch: master
Gerrit-MessageType: comment
Gerrit-Change-Id: Ia7ea6418f6190d7d0c6d5a1aaca4ca592c49ce41
Gerrit-Change-Number: 8048
Gerrit-PatchSet: 7
Gerrit-Owner: Andrew Wong <aw...@cloudera.com>
Gerrit-Reviewer: Adar Dembo <a...@cloudera.com>
Gerrit-Reviewer: Andrew Wong <aw...@cloudera.com>
Gerrit-Reviewer: Kudu Jenkins
Gerrit-Reviewer: Tidy Bot
Gerrit-Reviewer: Todd Lipcon <t...@apache.org>
Gerrit-Comment-Date: Thu, 05 Apr 2018 05:13:38 +0000
Gerrit-HasComments: Yes

Reply via email to