| 1 |
On Tue, 23 Nov 2004, Paul de Vrieze wrote: |
| 2 |
|
| 3 |
> On Tuesday 23 November 2004 09:38, Ed Grimm wrote: |
| 4 |
>> On Tue, 23 Nov 2004, Konstantin V. Gavrilenko wrote: |
| 5 |
>>> Guys, |
| 6 |
>>> |
| 7 |
>>> just thought I should mention this. |
| 8 |
>>> |
| 9 |
>>> if you setup a syslog-ng server that binds to an interface and listens |
| 10 |
>>> to collect incoming logs, at the next reboot it will fail to do so. I |
| 11 |
>>> believe this issue is due to the order of execution of the init.d |
| 12 |
>>> initialization scripts, since syslog-ng is started before the net.ethX |
| 13 |
>>> scripts are executed. The exim would not start following the failure |
| 14 |
>>> of the syslog-ng. |
| 15 |
>> |
| 16 |
>> My experience is, this is actually unpredictable - net.ethX does not |
| 17 |
>> provide 'net', and so whether syslog-ng (or, for that matter, any other |
| 18 |
>> syslog service) starts up before or after the network is established is |
| 19 |
>> anyone's guess. Given the total number of places it can start up on a |
| 20 |
>> typical system, I would guess that it probably usually starts after the |
| 21 |
>> network has initialized. |
| 22 |
> |
| 23 |
> Syslog-ng should, like iptables, bind to an interface even if it is not up. |
| 24 |
> There are reasons to start the logger as soon as possible. Alternatively you |
| 25 |
> could signal syslog-ng to reload itself again just before exim loads. |
| 26 |
|
| 27 |
Well, that's just great. Just after I thought I knew *exactly* what was |
| 28 |
going on, I managed to get a boot order that worked, that I thought |
| 29 |
wouldn't work. Syslog-ng started, followed by iptables, followed by |
| 30 |
net. Apparently, what was preventing syslog-ng from starting was not |
| 31 |
the absence of net. I'm going to investigate this one further. |
| 32 |
|
| 33 |
On the bright side, I have iptables starting reliably before net. So |
| 34 |
half of my problem appears to be solved. |
| 35 |
|
| 36 |
(For reference, I've had about 6 boots where syslog-ng failed, starting |
| 37 |
before net, and no failures where syslog-ng started fine, starting after |
| 38 |
net (around 10-15 boots). One boot where syslog-ng succeeded, starting |
| 39 |
before net. Syslog-ng last updated before I noticed the problem in the |
| 40 |
first place.) |
| 41 |
|
| 42 |
Ed |
| 43 |
|
| 44 |
-- |
| 45 |
gentoo-dev@g.o mailing list |
Archives