mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Adam B" <a...@mesosphere.io>
Subject Re: Review Request 41649: Removed slave checkpointing logic after deprecation cycle.
Date Sat, 09 Jan 2016 00:10:08 GMT


> On Jan. 6, 2016, 2:22 a.m., Joerg Schad wrote:
> > include/mesos/mesos.proto, line 482
> > <https://reviews.apache.org/r/41649/diff/3/?file=1175353#file1175353line482>
> >
> >     Let us doublecheck we actually need a deprecation cycle here: The critical scenario
would be in my opinion if a new slaveInfo (i.e. without) reaches an old master trying to access
this value.

We can address removing SlaveInfo.checkpoint in a separate patch. Although it has been forced
to true for a few(?) releases already, SlaveInfo is still part of the API for various hooks/decorators
as well as the executor [re]registered() APIs. Hooks must be compiled against the current
version, so they'll update safely. Unfortunately, older executors may still be compiled against
the old protobuf, which defaults 'checkpoint' to false if the field is missing. Perhaps we
should change the default value to 'true' during the deprecation cycle? I wonder if any executors
even care to look at the SlaveInfo.checkpoint field when they have MESOS_CHECKPOINT which
should describe whether the _framework_ checkpoints.
Also, as you mention, the slave reports its entire SlaveInfo to the master, which doesn't
act on SlaveInfo.checkpoint itself, but surfaces it in `/state` and `/slaves`, so external
scripts could theoretically consume the field.
I bet with an email to dev@/user@ we could clear this up and get (lazy) consensus to remove
the field from SlaveInfo, without a deprecation cycle.


- Adam


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/41649/#review112849
-----------------------------------------------------------


On Jan. 8, 2016, 4:46 a.m., Joerg Schad wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/41649/
> -----------------------------------------------------------
> 
> (Updated Jan. 8, 2016, 4:46 a.m.)
> 
> 
> Review request for mesos, Adam B, Joerg Schad, and Till Toenshoff.
> 
> 
> Bugs: MESOS-2317
>     https://issues.apache.org/jira/browse/MESOS-2317
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> Removed slave checkpointing logic after deprecation cycle.
> 
> 
> Diffs
> -----
> 
>   include/mesos/mesos.proto 158e08774c4a4fa5ec667388c61e55dbdafc7f67 
>   src/cli/execute.cpp a2b610f32da3deb9d3df99c225d22a425e03cdba 
>   src/examples/docker_no_executor_framework.cpp 7b3f0581f75105addb388dc15c4a0390ae4e0e0e

>   src/examples/java/TestFramework.java cbcdeedb0da22f82e274d088eb15a4128fd920fe 
>   src/examples/python/test_framework.py 6af6d22f54c0ad702917d09cf658e70f7032d793 
>   src/internal/devolve.cpp 6c38403d908954112e07030c80e8701b84178678 
>   src/master/allocator/mesos/hierarchical.hpp 86ea5a402ed67f8f22f11d5730147cd907d66a08

>   src/master/allocator/mesos/hierarchical.cpp df8bccaf2b8cfc0cb5ca18d4867371ae7a84c12f

>   src/master/master.cpp 40ce3e17fca88da689128bcf5d35fdddc396c011 
>   src/master/validation.cpp 6a43bce5b7df6a9d939245c4726d060fa19eb305 
>   src/slave/containerizer/containerizer.hpp 6964d136818ea9904fa35cd778eb9ef19e2c64fc

>   src/slave/flags.cpp a60d3c8022aba93fbd17a46dfff601fb1b25bbee 
>   src/slave/slave.cpp 9d80c96d8e28085c7fa47ce21b9b055c0926d12c 
>   src/slave/state.cpp b912c554fc9cb5593f541a7803422c22dc03dffd 
>   src/tests/slave_recovery_tests.cpp c0e4ff75b35c9e806741aab5696771e66d2c2ea8 
> 
> Diff: https://reviews.apache.org/r/41649/diff/
> 
> 
> Testing
> -------
> 
> make check
> 
> 
> Thanks,
> 
> Joerg Schad
> 
>


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