> On Jan. 23, 2015, 5:01 p.m., Adam B wrote:
> > include/mesos/fetcher/fetcher.proto, lines 44-45
> > <https://reviews.apache.org/r/30037/diff/6/?file=829684#file829684line44>
> >
> >     How does this fail for an item not present in the cache? Why wouldn't 
> > you just want to check for presence in the cache, retrieve if present and 
> > download&cache if absent?
> 
> Bernd Mathiske wrote:
>     Because I already know what's in the cache and what isn't. This 
> bookkeeping is NOT in mesos-fetcher, to which this is a parameter. It is in 
> FetcherProcess, which is in the slave.

Is there a writeup somewhere about what responsibilties fall in the realm of 
the FetcherProcess vs. mesos-fetcher? It would be interesting to know 
(especially as an implementer of an external/module fetcher) what state/logic 
belongs where, so that the API between the two is clear.


> On Jan. 23, 2015, 5:01 p.m., Adam B wrote:
> > include/mesos/fetcher/fetcher.proto, line 62
> > <https://reviews.apache.org/r/30037/diff/6/?file=829684#file829684line62>
> >
> >     Where did this go?
> 
> Bernd Mathiske wrote:
>     Turns out the env var HADOOP_HOME from which this was a carry-over is not 
> used anywhere in mesos-fetcher. It has been passed as an argument for ages 
> and not used at all. I checked back to April 2014. Nothing. So I removed it.

Have you tested your fetcher with HDFS fetching? Mesos assumes that the hdfs 
binary is either on the path or locatable from slave --hadoop_home or the 
HADOOP_HOME environment variable. See src/hdfs/hdfs.hpp for the HDFS() 
constructors. I don't think you need to make it a part of the 
MESOS_FETCHER_INFO protobuf/environment variable, but if the slave has a 
--hadoop_home set, that should be passed on to the fetcher in the HADOOP_HOME 
environment variable. This smells broken.


- Adam


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/30037/#review69514
-----------------------------------------------------------


On Jan. 25, 2015, 9:44 a.m., Bernd Mathiske wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/30037/
> -----------------------------------------------------------
> 
> (Updated Jan. 25, 2015, 9:44 a.m.)
> 
> 
> Review request for mesos, Adam B, Benjamin Hindman, Till Toenshoff, and 
> Timothy Chen.
> 
> 
> Bugs: MESOS-2069
>     https://issues.apache.org/jira/browse/MESOS-2069
> 
> 
> Repository: mesos-git
> 
> 
> Description
> -------
> 
> Extends the fetcher info protobuf with "actions" (fetch directly, through 
> cache, retrieve from cache). Switches the basis for dealing with uris to 
> "items", which contain the uri, the action, and potentially and cache file 
> name. Refactors fetch() and run(), so there is only one of each. Introduces 
> about half of the actual cache logic, including a hashmap of cache file 
> objects for bookkeeping and basic operations on it. 
> 
> 
> Diffs
> -----
> 
>   include/mesos/fetcher/fetcher.proto 
> facb87b92bf3194516f636dcc348e136af537721 
>   src/launcher/fetcher.cpp fed0105946da579a38357a30e7ae56e646e05b89 
>   src/slave/containerizer/fetcher.hpp 
> 1db0eaf002c8d0eaf4e0391858e61e0912b35829 
>   src/slave/containerizer/fetcher.cpp 
> d290f95251def3952c5ee34f600e1d71467f6293 
> 
> Diff: https://reviews.apache.org/r/30037/diff/
> 
> 
> Testing
> -------
> 
> make check
> 
> 
> Thanks,
> 
> Bernd Mathiske
> 
>

Reply via email to