asterixdb-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Till Westmann" <ti...@apache.org>
Subject Re: Documentation on NCService
Date Tue, 30 Jan 2018 07:13:04 GMT
Hi Steven,

AFAIK the NCService is intended to be run as a daemon on each node with the
sole purpose of managing the lifecycle of NCs as directed by the CC.
The lifecycle of the NCService itself should be managed by an operating
system service like sysvinit or systemd. Thus termination using a signal
seems appropriate - but SIGTERM (15) would to be a gentler choice than
SIGKILL (9).
The fact that SIGKILL doesn’t work for your process is a little more
confusing as it shouldn’t be possible for the process to ignore it ...

If you'd like to automate the installation on a cluster you could use
e.g. ansible [1] as described in our docs [2].

Does that help?

Cheers,
Till

[1] https://www.ansible.com/
[2] http://asterixdb.apache.org/docs/0.9.2/ansible.html

On 29 Jan 2018, at 13:59, Steven Jacobs wrote:

> Hi all,
> I'm trying to switch the BAD cluster from Managix to NCService, and it
> seems like the documentation is lacking. For example, there is
> documentation on starting the nc and cc service, but it seems like the only
> way to stop these things is by killing the services manually. Is that the
> only way that we have?
>
> I'm also seeing strange behavior where the ncservice processes are
> unkillable. I'm not sure whether this is some weird machine behavior or
> caused by ncservice (see below)
>
> Steven
>
> [sjacobs@dhcp-053228 three]$ jps
>
> 30871 NCService
>
> 31642 Jps
>
> 31023 NCService
>
> [sjacobs@dhcp-053228 three]$ kill -9 30871
>
> [sjacobs@dhcp-053228 three]$ jps
>
> 30871 NCService
>
> 31657 Jps
>
> 31023 NCService
>
> [sjacobs@dhcp-053228 three]$ kill `jps | egrep '(CDriver|NCService)' | awk
> '{print $1}'`
>
> [sjacobs@dhcp-053228 three]$ jps
>
> 30871 NCService
>
> 31690 Jps
>
> 31023 NCService

Mime
View raw message