Re: [Patch] tabled: parse Chunk parameters in CLD v2

[Jeff Garzik <jeff@garzik.org>]

Pete Zaitcev wrote:
> diff --git a/test/start-daemon b/test/start-daemon
> index 590da92..203e392 100755
> --- a/test/start-daemon
> +++ b/test/start-daemon
> @@ -20,10 +20,10 @@ fi
>  cld -d data/cld -P cld.pid -p 18081 -E
>  chunkd -C $top_srcdir/test/chunkd-test.conf -E
>  
> -# 3 is enough, but we like to let chunkd to come up eary and register with CLD
> +# 3 is enough, but we like to let chunkd to come up early and register with CLD
>  sleep 7
>  
> -../server/tabled -C $top_srcdir/test/tabled-test.conf -E
> +../server/tabled -C $top_srcdir/test/tabled-test.conf -E -D
>  
>  sleep 3
>  


Thinking out loud...

I occasionally wonder about a TCP control channel for tabled, to be used 
for various administration purposes.

tdbadm is not as useful when replication is active, unless tdbadm 
becomes the db4 master -- undesirable when the cluster is active.

An alternative, already employed by bind utility rndc(8), is to create 
an authenticated TCP-based control channel, to be used for various 
administrative purposes.

In tabled's case, we could use such a channel for

* database maintenance activities, while database cluster is active

* querying server startup status, so that you know _precisely_ when 
tabled is active.  That would permit you to eliminate the various "sleep 
3" and "sleep 7", I hope?

Regards,

	Jeff