Bugzilla – Bug 677314
option ntp-servers ignored, installer got stuck
Last modified: 2011-05-11 08:10:50 UTC
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:2.0b12) Gecko/20100101 Firefox/4.0b12 I provide "option ntp-servers" via dhcp (pointing to my local servers), but the installer ignored this, tried an external ntp resource instead, was blocked at the firewall and got stuck for several minutes. It should do as it is told. Reproducible: Always Steps to Reproduce: 1. 2. 3. Its the 11.4 rc2 net installer CD for amd64. I did not use automatic configuration.
PS: After clicking the [Finish] button at the end of the installation the boot procedure got stuck. Last words: sntp[xxxx]: Received no usable packet from 131.234.137.23. By now I waited for 10 minutes. Sorry, but this seems to be badly broken.
Well, under most circumstances it works pretty well - the problem is mostly caused by the firewall blocking access to the external time-server. As far as I know, we don't yet support setting the time-server from DHCP, although it would be nice to have. There are at least two openfate requests concerned with that.
Of course access to external services is restricted, esp. if the service can be found in my LAN. For Debian this was never a problem. Windows can handle this, too. Only openSUSE gets stuck (and old OpenBSD 4.3, but AFAICS they have fixed that). The NTP query should not get stuck at all, regardless whether the NTP server address was obtained via DHCP. I would suggest to reduce the timeout to 5 seconds. After the timeout it can continue trying in the background.
Same problem here. I've got an own ntp-server in the LAN and the machines I install don't have any access to the internet. I'm installing the machines via PXE in a separated network. The IPv4-configuration is provided with dhcp. The ntp-servers-option is ignored. Even more: if you disable the ntp-update during installation (after I waited very long time for the time out), the installed system wants to connect to the ntp-servers anyway. This takes about 17 minutes of time. It should be fixed very fast! It's a no go from my point of view - it's just pretty unusable! Is there a way to set the ntp-servers at the boot prompt like the install option?
This problem can be found in the production release, too.
Using the dhcp informations in the installer is a nice to have feature -> but file it at features.opensuse.org So if you disabled ntp during installation and it still contacts ntp servers, please provide your yast logs. And if you play with Priorities - assign it to yourself first, that field is for the developer.
This is feature request not a bug report
Hi Stephan and Peter, This is surely not a feature request. Using the default install the boot procedure still gets stuck for several minutes. This is not OK. As written above I would suggest to start ntpd in the background. Easy to fix. (I cannot remember having seen this problem with SLES, so maybe you might want to check their solution.) Sorry for being unresponsive before, but today I am back from vacation. I did not play with the Priorities. Regards Harri
1. There was never planned to use "option ntp-servers" during the installation. So it is a new feature. 2. In the installed system the "option ntp-servers" will be used, if you have configured ntp running as daemon and NETCONFIG_NTP_POLICY is configured resp. in /etc/sysconfig/network/config. 3. If the ntp server is not accessible this can cause timeouts. This is normal. How long this timeouts may be. We can discus about it. Unfortunately the default time out in sntp is very long and sntp have great powers of endurance, and tries the connections more time.