On 07/08/14 12:38, Jonathan Adams wrote: > DHCP address was not made unusable, when that does happen (it has happened, > when rebooting a machine sometimes it gets an echo from one of the > switches, now replaced, that lead it to think that the ip was taken) it was > just handed another address.
You don't get another address if you've configured the DHCP server for a static MAC address to IP address mapping (pntadm -f 03), which I believe was the original suggestion being discussed as an alternative to manually-configured static IP addresses on each network node. Under this failure condition, the address entry just gets marked "unusable" and you're toast because nothing else will be allocated (flag MANUAL 02). > In the case of this server, it got a response with an address but didn't > use it, just set the address to 0.0.0.0 ... this is the same Solaris DHCP > server that has worked non-stop since we moved into the building ~7 years > ago Hmm. I haven't seen that happen. Debug logs might be interesting. In any event, on my home system, I've had to disable the Solaris DHCP server and switch to ISC due to reliability problems. The Solaris DHCP server seems to get stuck all too often like this: Jun 8 15:47:33 carlson in.dhcpd[25002]: [ID 583682 daemon.error] SENDTO: Network is unreachable. Jun 8 15:53:24 carlson last message repeated 2 times Jun 8 15:56:33 carlson in.dhcpd[25002]: [ID 583682 daemon.error] SENDTO: Network is unreachable. Jun 8 16:08:36 carlson last message repeated 3 times Jun 8 16:15:00 carlson in.dhcpd[25002]: [ID 583682 daemon.error] SENDTO: Network is unreachable. Jun 8 16:21:27 carlson last message repeated 2 times The error messages (and the truss output) are completely nonsensical, because the server's interface in question is up, running, and working fine on its own. Something just plain comes unglued in the plain IP (non-DLPIv2) interfaces that in.dhcpd is using. When that happens, nobody gets addresses, and everyone in the house starts hounding me about the "network problems." I tried to track it down but just gave up. I spent a few months with a nasty crontab entry that did "svcadm restart dhcp-server" daily, but even that wasn't enough. Since switching to ISC a month ago, I haven't had outages like that. (Yeah, I don't feel good about it, having previously worked in the group that maintained this stuff. But life moves on.) > We have had problems in the past at another site, with Windows PC's on the > Solaris DHCP server (we put it down to the wiring, but it could be > anything) that PC's running Windows XP and Microsoft office, at the point > in time that the lease is renegotiated would reconnect to the samba server > without first disconnecting, causing the document to deny writing because > it was already open for editing, and automatically closing ... in that case > setting static IP was the easiest solution. I haven't run into that. I normally get back the same address on lease renewal. I'd certainly call it a server bug if you *don't* get the same address on renewal and you haven't also completely run out of addresses. -- James Carlson 42.703N 71.076W <carls...@workingcode.com> _______________________________________________ openindiana-discuss mailing list openindiana-discuss@openindiana.org http://openindiana.org/mailman/listinfo/openindiana-discuss
