mesos-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Cong Wang" <cw...@twopensource.com>
Subject Re: Review Request 38473: Add flag to disable hostname lookup.
Date Mon, 21 Sep 2015 18:09:25 GMT


> On Sept. 18, 2015, 10:57 p.m., Cong Wang wrote:
> > Why not just set --host_name=$LIBPROCESS_IP for your case since you anyway need
to provide a flag?
> 
> Guangya Liu wrote:
>     I also have the same question with Cong, @Marco, can you please show more detail
for why not using the solution as above? What is the benefit of this compared to using --host_name=$LIBPROCESS_IP?
Thanks.
> 
> Marco Massenzio wrote:
>     `$LIBPROCESS_IP` may not be set by the time we run the binary - it is actually set
in the `main()` method (of both Master/Agent) after parsing the `--ip` (and `--ip_discovery_command`)
flags.
>     
>     The reason I use `LIBPROCESS_IP` here is to ensure that the IP used by `libprocess`
(which is initialized before creating the `Master` or `Slave` object) is consistent with what
the `Master/Agent` will be configured with.
>     
>     Please also refer to the conversation about the introduction of `--ip_discovery_command`
as to why we need a "dynamic" way of configuring hostname/IP - it all boils down to make Mesos
play nice in various Cloud (and Enterprise data centers) where the DNS resolution (and hostname
configuration of VMs) is vastly different and may not necessarily conform to the generally
assumed scenario, where IP/hostname are kept in sync / resolved by the OS and/or external
services (DHCP/DNS).
> 
> Qian Zhang wrote:
>     If you want to set hostname to the chosen IP, then I think you can just use the existing
flag "--hostname" (set it to the chosen IP manually). And if "--hostname" is set, then I think
the hostname lookup will be bypassed. So why do we need to introduce a new flag "--no_hostname_lookup"?
> 
> Marco Massenzio wrote:
>     Please refer to the conversation on https://issues.apache.org/jira/browse/MESOS-3154
for why this may not be possible in the above-mentioned environments.
>     (and, yes, we do know about the `--hostname` flag - unfortunately, under certain
scenarios, it's not possible to set it to a "static" value)

Makes sense for me, can you document it in configuration.md too? Thanks.


- Cong


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/38473/#review99628
-----------------------------------------------------------


On Sept. 19, 2015, 1:25 a.m., Marco Massenzio wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/38473/
> -----------------------------------------------------------
> 
> (Updated Sept. 19, 2015, 1:25 a.m.)
> 
> 
> Review request for mesos, Benjamin Hindman, Cody Maloney, and Neil Conway.
> 
> 
> Bugs: MESOS-3457
>     https://issues.apache.org/jira/browse/MESOS-3457
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> Under certain circumstances, dynamic lookup of hostname, while
> successful, provides undesirable results; we would also like, in
> those circumstances, be able to just set the hostname to the chosen
> IP address (possibly set via the --ip_discovery_command method).
> 
> The flag we add here, --[no]-hostname_lookup is `true` by default
> (which is the existing behavior) and will work under most
> circumstances: however, by disabling lookup (using, for example,
> --no_hostname_lookup) the hostname used will be the same as the
> one configured (possibly, via --ip or MESOS_IP) in `LIBPROCESS_IP`.
> 
> This change affects both Master/Agent nodes.
> 
> WARNING - the `Address::hostname()` method always does a dynamic
> hostname lookup, which may in fact return inconsistent results
> wrt the Master's configuration (this is *not* affected by
> this change) and should be avoided; use instead
> `Master::info()::hostname()` which is always consistent with
> the Master's configuration.
> 
> 
> Diffs
> -----
> 
>   docs/configuration.md dd7f4aa806a3c1a8653a0fda9a326a3707308e7c 
>   src/master/flags.hpp e4b1df3f5a33049defff4688463274067f1f1ebf 
>   src/master/flags.cpp 80879611fbcfd764c9fc8f60a31613a9c8fc2364 
>   src/master/master.cpp ca4d5876dcd427964111428edc22d567ddaede0b 
>   src/slave/flags.hpp e31a4183170c3442ac4a15365c229391e7e91480 
>   src/slave/flags.cpp add4196dfd06c0f602ff5ebd39960dc05c4cd11f 
>   src/slave/slave.cpp ad710d7b930a2f115d503ceb8f8fd7421ad30287 
>   src/tests/cluster.hpp 114583de8c867495a2b7a953e6826708838e5d23 
>   src/tests/master_tests.cpp a477794df37c658b80d79dea8555b001415f7b6a 
>   src/tests/mesos.hpp 3db97aca921c9216d90384e1eb17030849516454 
> 
> Diff: https://reviews.apache.org/r/38473/diff/
> 
> 
> Testing
> -------
> 
> make check
> 
> 
> Thanks,
> 
> Marco Massenzio
> 
>


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