Slony1 PostgreSQL Replication - rebuildlistenentries for enabled nodes only

Author

rebuildlistenentries for enabled nodes only

Andreas Pflug

2005-09-14, 7:23 am

We've been discussing using nodes/paths to store information for
non-slon active node, but didn't come to a conclusion so far.

Currently, listens will be generated for every path regardless of the
node's activity. If a node will never have a running slon daemon, events
for this node will collect and prevent cleanup of old events, which is bad.

As a fix, I'd propose to generate listens from paths only between
_active_ nodes, i.e. add "and N1.no_active and N2.no_active" to the
select condition in rebuildlistenentries
(). A node that's not active
will never process events, and never generate events so IMHO it's
correct to have no listens for such nodes. In addition,
rebuildlistenentries
() probably needs to be added to enablenode() and
disablenode() (as soon as it's implemented) too.

Regards,
Andreas

Andreas Pflug

2005-09-19, 11:24 am

May I plase have a little feedback on this?
pgsql 8.1 release is near, so we'll have to release pgAdmin III V1.4
quite soon too. I'd really like to have correct slony support in it too;
I'll have finish the work in the next very few weeks.

Regards,
Andreas

Andreas Pflug wrote:

> We've been discussing using nodes/paths to store information for
> non-slon active node, but didn't come to a conclusion so far.
>
> Currently, listens will be generated for every path regardless of the
> node's activity. If a node will never have a running slon daemon,
> events for this node will collect and prevent cleanup of old events,
> which is bad.
>
> As a fix, I'd propose to generate listens from paths only between
> _active_ nodes, i.e. add "and N1.no_active and N2.no_active" to the
> select condition in rebuildlistenentries
(). A node that's not active
> will never process events, and never generate events so IMHO it's
> correct to have no listens for such nodes. In addition,
> rebuildlistenentries
() probably needs to be added to enablenode() and
> disablenode() (as soon as it's implemented) too.
>
> Regards,
> Andreas
> ____________________
____________________
_______
> Slony1-general mailing list
> Slony1-general- AuKwsB3Fm+ugFIWk8tvy
RWD2FQJk+8+b@public.gmane.org
> http://gborg.postgresql.org/mailman.../slony1-general

Andreas Pflug

2005-09-25, 7:23 am

Attached a patch implementing my proposal.

Regards,
Andreas

Andreas Pflug wrote:
> We've been discussing using nodes/paths to store information for
> non-slon active node, but didn't come to a conclusion so far.
>
> Currently, listens will be generated for every path regardless of the
> node's activity. If a node will never have a running slon daemon, events
> for this node will collect and prevent cleanup of old events, which is bad.
>
> As a fix, I'd propose to generate listens from paths only between
> _active_ nodes, i.e. add "and N1.no_active and N2.no_active" to the
> select condition in rebuildlistenentries
(). A node that's not active
> will never process events, and never generate events so IMHO it's
> correct to have no listens for such nodes. In addition,
> rebuildlistenentries
() probably needs to be added to enablenode() and
> disablenode() (as soon as it's implemented) too.
>
> Regards,
> Andreas
> ____________________
____________________
_______
> Slony1-general mailing list
> Slony1-general- AuKwsB3Fm+ugFIWk8tvy
RWD2FQJk+8+b@public.gmane.org
> http://gborg.postgresql.org/mailman.../slony1-general

Dave Page

2005-09-25, 8:24 pm

If there are no objections I'll apply this to head and the 1.1 branch in
the next day or so.

Regards, Dave.=20

> -----Original Message-----
> From: slony1-general-bounces- AuKwsB3Fm+ugFIWk8tvy
RWD2FQJk+8+b@public.gmane.org=20
> [mailto:slony1-general-bounces- AuKwsB3Fm+ugFIWk8tvy
RWD2FQJk+8+b@public.gmane.org] On=20
> Behalf Of Andreas Pflug
> Sent: 25 September 2005 11:54
> Cc: Slony Mailing List
> Subject: Re: [Slony1-general] rebuildlistenentries
for=20
> enabled nodes only
>=20
> Attached a patch implementing my proposal.
>=20
> Regards,
> Andreas
>=20
> Andreas Pflug wrote:
> regardless of the=20
> daemon, events=20
> events, which is bad.
> not active=20
> enablenode() and=20
>=20
>=20

Sponsored Links