Topic: SME 5.5 stopping boot on NTPD

[Jáder Marasca]

Hi,

I´m installing a new SME 5.5 on a K6II 400 with MVP4 chipset and 256RAM and HDD 20G.
Everything else is normal. BUT when I press the power button to start computer , some times SME stop to load ALLWAYS at same point: NTPD...
So I press RESET and after check HDD for errors it starts ok...

Not a BIG problem if my client left it turned on (I still don´t know if it will!) but if they will start over each day... I´ll have a serious problem.

Anyone has ANY idea WHY this SH**T is happening?

Thanks!

Jáder

[Robert]

No need to get rude
On startup ntpd tries to sync with the time server. If your internet connection is through ppp or pppoe this will fail because the external interface is not up yet. Ntpd should time out gracefully if that is the case (may take half a minute or so). If it doesn't, look in the messages log what went wrong. Possible workarounds might be to disable ntp or to move it to later in the startup sequence.

[Jáder Marasca]

Thanks Robert,

Robert wrote:
>Possible workarounds might be to disable ntp or
> to move it to later in the startup sequence.

I´ll try to move it to later... I assume this is done on rc.7 link with higher number. Isn´t?

[Robert]

You must also change the InitscriptOrder property in the configuration file or a restart by esmith serviceControl will fail. So, you could do:
# mv /etc/rc.d/rc7.d/S{55,90}ntpd
# /sbin/e-smith/config setprop ntpd InitscriptOrder 90

[Jáder Marasca]

Thanks by the info!

I had changed it to 77 (just the mv command!) and hadn´t solved!

I´ll try to shutdown/restart server at 10 times today to verify if this is ok now!

I´ll install it tomorrow at client site!

Thanks!

Jáder

[Robert]

I thought about this some more and found a potential problem with moving the ntpd initscript to later in the startup sequence which could have you end up with two ntpd instances.
Connecting through ppp or pppoe will trigger the ip-change event, which includes S90restart-ntpd. This will stop and start ntpd. If ntpd wasn't running, stop will fail, but start will succeed. If /etc/e-smith/events/ip-change/S90restart-ntpd finishes before /etc/rc.d/rc7.d/SXXntpd is called, you'll have two ntpd instances. I suppose you could fix this by making ntpd startup conditional on the absence of the ntpd lockfile by editing /etc/rc.d/init.d/ntpd (make a copy!) and changing
start)
     start
to
start)
     [ -f /var/lock/subsys/ntpd ] || start
, but that would be yet another workaround without fixing or identifying the real problem, which may or may not be that ntpdate hangs instead of timing out. What do the logs tell you about what the real problem might be?

[Jáder Marasca]

Robert

Thanks by the info... I think I´ll let NTP at 90... It don´t stop the error... but it happens a few times... and I´ll never have a IP change (fixed IP on this computer!) so have NTPd at 90 will not be an issue!

BTW I have TWO (of so many shutdown & power up cycles) where SME stopped at SMB.conf and not a NTPd as usual...

I´m getting confused... I NEVER saw SME stop at loading process... and once it loads everything is fine!

Thanks by your help so far!

Jáder

[Robert]

I've only had my system hang once. That was because of a faulty ethernet driver. Most system hangs seem to be related to faulty hardware. Are you sure there is nothing in the messages logs that could explain these mysterious hangs?