mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Chun-Hung Hsiao <chhs...@apache.org>
Subject Re: Review Request 68601: Cleaned up residual CSI endpoint sockets for terminated plugins.
Date Mon, 03 Sep 2018 22:23:05 GMT


> On Sept. 3, 2018, 10:52 a.m., Benjamin Bannier wrote:
> > src/resource_provider/storage/provider.cpp
> > Lines 701 (patched)
> > <https://reviews.apache.org/r/68601/diff/2/?file=2081301#file2081301line701>
> >
> >     I'd pull this down into the context making use of the value. Otherwise it isn't
very clear that `nodeContainerId` doesn't become unset between check and use.

The function actually no longer uses this pre-condition so I'll just remove it.


> On Sept. 3, 2018, 10:52 a.m., Benjamin Bannier wrote:
> > src/resource_provider/storage/provider.cpp
> > Line 770 (original), 805 (patched)
> > <https://reviews.apache.org/r/68601/diff/2/?file=2081301#file2081301line807>
> >
> >     Not your patch, but are we guaranteed to always remove the directory? What it
the agent crashes between the `killContainer` call and this line?

The `endpointDir` is symlinked from `<containerPath>/endpoint`, so during SLRP recovery
it will be picked up and removed.


> On Sept. 3, 2018, 10:52 a.m., Benjamin Bannier wrote:
> > src/resource_provider/storage/provider.cpp
> > Line 2093 (original), 2163 (patched)
> > <https://reviews.apache.org/r/68601/diff/2/?file=2081301#file2081301line2193>
> >
> >     I am not sure how this is related. Is this a patch 4?

Actually the `WAIT_CONTAINER` call should not return a `Not Found` because `KILL_CONTAINER`
didn't return a `Not Found`. I could put up another patch to fix this.


- Chun-Hung


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/68601/#review208250
-----------------------------------------------------------


On Sept. 3, 2018, 12:44 a.m., Chun-Hung Hsiao wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/68601/
> -----------------------------------------------------------
> 
> (Updated Sept. 3, 2018, 12:44 a.m.)
> 
> 
> Review request for mesos, Benjamin Bannier and Jie Yu.
> 
> 
> Bugs: MESOS-8429
>     https://issues.apache.org/jira/browse/MESOS-8429
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> If a CSI plugin is crashed during agent failover, the residual socket
> file would exist during SLRP recovery, which may in turn make the plugin
> fail to restart. This patch cleans up the residual socket files to avoid
> such failures.
> 
> 
> Diffs
> -----
> 
>   src/authorizer/local/authorizer.cpp f99b88e10df1e0959f1ddd2e45374862c2dc0a5b 
>   src/internal/devolve.hpp 1bc2a32e70600c785660d0b1d5d63378ffe344cc 
>   src/internal/devolve.cpp 93bd975b648888a9d83e34f4d7420b3b49f06d1f 
>   src/resource_provider/storage/provider.cpp ab1467c22ded269cb42fc52d3f00fb936fc42c7f

>   src/slave/http.cpp d0f6e1c5ffcc22cd280c64b20473f279ab4c14cc 
> 
> 
> Diff: https://reviews.apache.org/r/68601/diff/2/
> 
> 
> Testing
> -------
> 
> sudo make check
> 
> 
> Thanks,
> 
> Chun-Hung Hsiao
> 
>


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