Topic: dhcpd restarts - why?

[Lasse Johansson]

I have found something rather unpleasant:

When I create a new project place for the staff, I usually first design the group of people (users) and then create the actual ibay.
This is what happens:

1. Creating the group (in this example "project3223"), all normal. smbd restarts which seems perfectly OK.

2. Creating the ibay (in this example "proj3223") and assigning write permissions to the new group ("project3223"), other users read access, the following occurs:

*************************************************************************************
Sep  4 14:27:17 intranet2 e-smith[19114]: Processing event: ibay-create proj3223
Sep  4 14:27:17 intranet2 e-smith[19114]: Running event handler: /etc/e-smith/events/ibay-create/S15ibay-create
Sep  4 14:27:17 intranet2 useradd[19117]: new group: name=proj3223, gid=586
Sep  4 14:27:17 intranet2 useradd[19117]: new user: name=proj3223, uid=568, gid=586, home=/home/e-smith/files/ibays/proj3223/files, shell=/bin/false
Sep  4 14:27:17 intranet2 useradd[19117]: add proj3223' to group shared'
Sep  4 14:27:17 intranet2 useradd[19117]: add proj3223' to shadow group shared'
Sep  4 14:27:18 intranet2 PAM_pwdb[19124]: password for (proj3223/568) changed by ((null)/0)
Sep  4 14:27:18 intranet2 e-smith[19114]: Running event handler: /etc/e-smith/events/ibay-create/S55conf-servers
Sep  4 14:27:19 intranet2 e-smith[19114]: Running event handler: /etc/e-smith/events/ibay-create/S90restart-servers
Sep  4 14:27:19 intranet2 modprobe: can't locate module net-pf-10
Sep  4 14:27:19 intranet2 last message repeated 2 times
Sep  4 14:27:19 intranet2 named[559]: reloading nameserver
Sep  4 14:27:19 intranet2 named[559]: Forwarding source address is [0.0.0.0].1885
Sep  4 14:27:19 intranet2 named[559]: Ready to answer queries.
Sep  4 14:27:20 intranet2 modprobe: can't locate module net-pf-10
Sep  4 14:27:20 intranet2 afpd[19108]: shutting down on signal 15
Sep  4 14:27:20 intranet2 modprobe: can't locate module net-pf-10
Sep  4 14:27:20 intranet2 modprobe: can't locate module net-pf-10
Sep  4 14:27:20 intranet2 rlinetd[340]: rlinetd configuration (re)loaded, -32 bytes used
Sep  4 14:27:20 intranet2 atalkd[19004]: done
Sep  4 14:27:21 intranet2 xntpd[19006]: xntpd exiting on signal 15
Sep  4 14:27:21 intranet2 smb: smbd shutdown succeeded
Sep  4 14:27:21 intranet2 lpd: lpd shutdown succeeded
Sep  4 14:27:21 intranet2 dhcpd: dhcpd shutdown succeeded
Sep  4 14:27:21 intranet2 xntpd: xntpd shutdown succeeded
Sep  4 14:27:22 intranet2 atalkd[19276]: restart (1.4b2+asun2.1.3)
Sep  4 14:27:22 intranet2 lpd: lpd startup succeeded
Sep  4 14:27:22 intranet2 smb: nmbd shutdown succeeded
Sep  4 14:27:22 intranet2 xntpd[19308]: xntpd 3-5.93e Mon Aug 16 20:55:26 EDT 1999 (1)
Sep  4 14:27:22 intranet2 xntpd[19308]: tickadj = 5, tick = 10000, tvu_maxslew = 495, est. hz = 100
Sep  4 14:27:22 intranet2 xntpd[19308]: precision = 10 usec
Sep  4 14:27:22 intranet2 xntpd[19308]: read drift of 34.717 from /etc/ntp/drift
Sep  4 14:27:22 intranet2 xntpd[19308]: bind() fd 11, family 2, port 123, addr e0000101, in_classd=1 flags=0 fails: Address already in use
Sep  4 14:27:22 intranet2 xntpd[19308]: ...multicast address 224.0.1.1 using wildcard socket
Sep  4 14:27:22 intranet2 xntpd: xntpd startup succeeded
Sep  4 14:27:22 intranet2 dhcpd: Internet Software Consortium DHCP Server 2.0
Sep  4 14:27:22 intranet2 dhcpd: Copyright 1995, 1996, 1997, 1998, 1999 The Internet Software Consortium.
Sep  4 14:27:22 intranet2 dhcpd: All rights reserved.
Sep  4 14:27:22 intranet2 dhcpd:
Sep  4 14:27:22 intranet2 dhcpd: Please contribute if you find this software useful.
Sep  4 14:27:22 intranet2 dhcpd: For info, please visit http://www.isc.org/dhcp-contrib.html
Sep  4 14:27:22 intranet2 dhcpd:
Sep  4 14:27:22 intranet2 dhcpd: Internet Software Consortium DHCP Server 2.0
Sep  4 14:27:22 intranet2 dhcpd: Copyright 1995, 1996, 1997, 1998, 1999 The Internet Software Consortium.
Sep  4 14:27:22 intranet2 dhcpd: All rights reserved.
Sep  4 14:27:22 intranet2 dhcpd:
Sep  4 14:27:22 intranet2 dhcpd: Please contribute if you find this software useful.
Sep  4 14:27:22 intranet2 dhcpd: For info, please visit http://www.isc.org/dhcp-contrib.html
Sep  4 14:27:22 intranet2 dhcpd:
Sep  4 14:27:22 intranet2 dhcpd: Listening on LPF/eth0/00:10:4b:11:0c:ef/192.6.1.0
Sep  4 14:27:22 intranet2 dhcpd: Listening on LPF/eth0/00:10:4b:11:0c:ef/192.6.1.0
Sep  4 14:27:22 intranet2 dhcpd: Sending on   LPF/eth0/00:10:4b:11:0c:ef/192.6.1.0
Sep  4 14:27:22 intranet2 dhcpd: Sending on   Socket/fallback/fallback-net
Sep  4 14:27:22 intranet2 dhcpd:
Sep  4 14:27:22 intranet2 dhcpd: Sending on   LPF/eth0/00:10:4b:11:0c:ef/192.6.1.0
Sep  4 14:27:23 intranet2 dhcpd: Sending on   Socket/fallback/fallback-net
Sep  4 14:27:23 intranet2 dhcpd: dhcpd startup succeeded
*************************************************************************************

Why does dhcpd have to restart?
What does the line
"Running event handler: /etc/e-smith/events/ibay-create/S90restart-servers"
include, and why?

The result is that lots of workstations on the LAN loses connection with the server and have to restart...

Any of you E-Smith-gurus out there who have any clues????

//Lasse

[Damien Curtain]

dhcpd restarting wont effect your workstations unless at that exact point theyre renewing their leases, but luckily they renew leases a short while before absolutely neccessary to ensure they can cope with situations like the above, so only if a new lease is sought at that exact point will it wait and retry.

restart-servers restarts servers. Thats what it does. When you add a new ibay several services require reconfiguration, your right that dhcpd doesnt, smb infact will notice thats its config file has changed and not require a restart (someone very wise pointed this out to me),  named just needs a reload, etc... definately a finer grained solution is on its way. restart-servers is currently the all encompassing way to ensure the servers running on e-smith reread their (updated) settings.

less /etc/e-smith/events/ibay-create/S90restart-servers will tell you exactly whats being done

--
 Damien

[Lasse Johansson]

Thank you for your answer, - but it really doesn't solve my problem!

If what you say is true (probably is!), then something else is really screwed up, because, as I wrote in my first report:

"The result is that lots of workstations on the LAN loses connection with the server and have to restart..."

...That includes my own workstation as well. And more: It's not even enough to log out and log in again... Without actually RESTARTING the client, windows informs that "there is no such domain available"... (Yes, we use netlogon)

That's why I tried to scan through /var/log/messages for clues...

So, where else should I look?

The network is a HEAVILY LOADED network of a busy research institute, filled with scientists and doctors, all together between 35-40 users. We mainly use the E-Smith box as file server (server only mode) but also as PDC and print server for our 4 network printers. I'm really hesitant to implement new ibays during work hours now after this problem ( - something I used to do frequently earlier without any problems)

We also have other E-Smith boxes on the net for other purposes, and we have no problems at all with those.

Best Regards

//Lasse

[Gordon Rowell]

Lasse Johansson wrote:

> Thank you for your answer, - but it really doesn't solve my
> problem!
> [...]

I think Damien spotted it. restart-servers currently performs
the command "smb restart". That shouldn't be required in these
cases - "smb reload" is sufficient (and also probably not required).

I'll take this up off-line and then post results to the forum once
we've determined what's happening.

Gordon

[Gordon Rowell]

Gordon Rowell wrote:

> [...]
> I think Damien spotted it. restart-servers currently performs
> the command "smb restart". That shouldn't be required
> in these cases - "smb reload" is sufficient (and also
> probably not required).
>
> I'll take this up off-line and then post results to the forum
> once we've determined what's happening.

And it was as Damien spotted. Changing conf-servers so that it does
"smb reload" instead of "smb restart"  fixes the problem. The latter
may not even be required as Samba notices when smb.conf is changed.

Gordon