Re: squid, DNS and no buffer space
Stephen Borrill <[email protected]>
| Newsgroups | gmane.os.netbsd.devel.network |
|---|---|
| Message-ID | <Pine.NEB.4.64.2412030904240.1680@ugly> |
On Tue, 3 Dec 2024, Lloyd Parkes wrote: > On Mon, 2024-12-02 at 15:11 +0000, Stephen Borrill wrote: >> On a busy squid proxy server, users are seeing performance problems >> and >> squid is logging the following: >> >> 2024/12/02 15:00:06 kid1| comm_udp_sendto FD 9, (family=2) >> 127.0.0.1:53: (55) No buffer space available >> 2024/12/02 15:00:06 kid1| idnsSendQuery FD 9: sendto: (55) No buffer >> space available >> > > Now I remember, I have seen this before. BIND assumes it's running as a > forwarder for an ISP and that it will therefore be expected to cache > the entire Internet's DNS. It will use all available memory. > > I had to add the following > > max-cache-size 25%; > > to the options block in my named.conf to get reasonable behaviour on my > multi-purpose server. I did look at top at the time to see if named was being particularly greedy, but didn't spot anything. I will look more closely next time, but interactively I don't recall any delay doing DNS lookups. Changing that setting is a good shout though. One of my co-workers was using our internal resursive DNS to clean a huge domain blacklist and hit exactly the problem you describe. -- Stephen