Re: [PATCH] dnscache: merge similar outgoing udp packets
Mark Johnson <[email protected]>
| Newsgroups | gmane.network.djbdns |
|---|---|
| Message-ID | <[email protected]> |
On Tue, Apr 21, 2009 at 3:10 PM, Jeff King <[email protected]> wrote: > On Thu, Apr 16, 2009 at 05:26:40PM -0500, Mark Johnson wrote: > >> > This patch is only lightly tested. Use on production servers at your own >> > risk (and please report to the list if you have success using it). >> >> This will be going into the much delayed zinq-djbdns-0.06. > > Thanks. The list has been silent on success reports, so perhaps nobody > is actually using it. I've been running it for a few weeks now without > incident, but only on a lightly loaded server (and only on dnscache -- > the changes shouldn't affect other programs, but I haven't tested it > extensively). I ran it for a few days before pushing it out and observed no problems (and received no reports of problems) with dnscache or tinydns. I don't know that zinq-djbdns actually has any users (as opposed to downloaders), but if I get any trouble reports, I'll let you know. >> Recommended value for MERGEQUERIES? Obviously dnscache doesn't care, >> but envdir will not set an environment variable if the corresponding >> file is empty. I'd be tempted to just use '1' or 'ON' except that >> suggests you can turn query merging off by setting it to '0' or 'OFF', >> when what they really have to do is remove the file or truncate it to >> zero bytes. Is there a best practice? I do intend to modify >> dnscache-conf to create a MERGEQUERIES file in the env directory so >> that query merging is on by default in zinq-djbdns. > > I would just use "1", which I see you did in the new zinc release. I > agree that the "0 still enables" behavior is confusing and has bitten > people on the list, but I think it is better to be consistent with the > existing options (which is why I implemented it as I did). > > -Peff I slapped a note in INSTALL about disabling query merging. Suggestions for better documentation always appreciated.