mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From James Peach <jpe...@apache.org>
Subject Re: Review Request 55795: Fixed `umountAll` to optionally also clean up mtab.
Date Tue, 24 Jan 2017 23:55:27 GMT


> On Jan. 23, 2017, 9:39 p.m., James Peach wrote:
> > src/linux/fs.cpp, line 414
> > <https://reviews.apache.org/r/55795/diff/1/?file=1611031#file1611031line414>
> >
> >     Why do you need `cleanUpMtab`. All the callers pass `true` and it is pretty
hard to imagine a scenario where you *want* the mtab to be out of sync.
> 
> Jiang Yan Xu wrote:
>     Yeah I don't think anyone would want that but if the caller mounts everything exclusively
using the syscall under `target` then they don't *need* to clean up mtab. (There's a minor
cost to it and potential failure factors but admittedly I haven't seen it fail in cases including
non-existent /etc/mtab, non-existent entry, etc.)
>     
>     So maybe it's fine to leave out the boolean argument and fix things later if anything
breaks in any environments.
> 
> James Peach wrote:
>     This is already subject to partial failure, but I'd also be OK with ignoring the
return from executing `umount`, or refactoring this to rewrite `mtab` using `setmntent(3)`.
I think that this patch is an existence proof that no-one really knows they don't need to
clean up :)
> 
> Jiang Yan Xu wrote:
>     re `setmntent(3)`: sounds worthy of a TODO but given that they are currently used
mainly for tests I would take this short cut for now :)
>     re partial failure: I guess it's unavoidable. I don't think this method itself knows
if it should ignore the failure, right now it returns an error when `fs::unmount` fails on
a mount entry so I guess returning an error when `umount --fake` fails is basically treating
the two as two equal steps taken for each mount.

Yup I agree. Leave it as is but remove the `cleanUpMtab` arg?


- James


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/55795/#review162706
-----------------------------------------------------------


On Jan. 20, 2017, 11:32 p.m., Jiang Yan Xu wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/55795/
> -----------------------------------------------------------
> 
> (Updated Jan. 20, 2017, 11:32 p.m.)
> 
> 
> Review request for mesos, Gilbert Song, Jie Yu, and James Peach.
> 
> 
> Bugs: MESOS-6478
>     https://issues.apache.org/jira/browse/MESOS-6478
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> - Without this, Mesos tests leave garbage in /etc/mtab.
> 
> 
> Diffs
> -----
> 
>   src/linux/fs.hpp da49c9ebfa938d169152ed3b6e4df7378711b013 
>   src/linux/fs.cpp 913e23317291db164fe6bdf77f3eca146dedec9b 
>   src/tests/containerizer/docker_volume_isolator_tests.cpp 4040d36d1dbf4968f62fa593024936858f28b4df

>   src/tests/environment.cpp a9e217e8beb5a6aca456ff5379893953cafca135 
>   src/tests/persistent_volume_tests.cpp 468a85b4a6ce09592341afd07ce12a03f5fc4f73 
> 
> Diff: https://reviews.apache.org/r/55795/diff/
> 
> 
> Testing
> -------
> 
> sudo make check.
> 
> Verified that no garbage is left in mtab.
> 
> 
> Thanks,
> 
> Jiang Yan Xu
> 
>


Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message