[opensuse] Postfix and clock weirdness.
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hi, I just have seen something in the warn log and I'm investigating. See the detailed log: Feb 12 12:26:54 nimrodel postfix/local[29204]: ED7FE86F23: to=<cer@localhost.nimrodel.valinor>, relay=local, delay=62, status=sent (delivered to command: /usr/bin/procmail) Feb 12 12:26:54 nimrodel postfix/smtpd[29128]: connect from localhost[127.0.0.1] Feb 12 12:26:52 nimrodel postfix/smtpd[29126]: disconnect from localhost[127.0.0.1] Feb 12 12:26:52 nimrodel amavis[26348]: (26348-06-28) FWD via SMTP: <opensuse-es+bounces-6521-robin.listas... Notice the clock jumping back two whole seconds? Well, postfix gives notice: Feb 12 12:26:52 nimrodel postfix/qmgr[4707]: warning: backward time jump detected -- slewing clock ... Feb 12 12:27:34 nimrodel postfix/qmgr[4707]: warning: backward time jump recovered -- back to normality Now, why did the clock jump back two whole seconds? Is postfix adjusting the clock, or is it just noticing it? I'm using ntp, it shouldn't do that! But it does. See the xntpd log: 12 Feb 03:02:53 ntpd[4635]: offset -0.003808 sec freq 10.337 ppm error 0.010344 poll 10 12 Feb 03:03:28 ntpd[4635]: synchronized to 193.127.101.30, stratum 2 (here the system was suspended to disk, thus the lap, from 03:39 to 11:34) 12 Feb 11:35:45 ntpd[4635]: synchronized to 194.146.227.112, stratum 2 12 Feb 11:39:39 ntpd[4635]: synchronized to 193.225.14.163, stratum 2 12 Feb 11:49:13 ntpd[4635]: synchronized to LOCAL(0), stratum 10 12 Feb 11:52:47 ntpd[4635]: synchronized to 194.150.121.14, stratum 2 12 Feb 11:57:24 ntpd[4635]: offset 0.000000 sec freq 10.363 ppm error 0.122509 poll 10 12 Feb 12:26:52 ntpd[4635]: time reset -2.125240 s <====== 12 Feb 12:26:52 ntpd[4635]: system event 'event_clock_reset' (0x05) status 'leap_none, sync_unspec, 15 events, event_peer/strat_chg' (0xf4) 12 Feb 12:26:52 ntpd[4635]: system event 'event_peer/strat_chg' (0x04) status 'leap_none, sync_unspec, 15 events, event_clock_reset' (0xf5) 12 Feb 12:26:54 ntpd[4635]: peer 193.127.101.30 event 'event_reach' (0x84) status 'unreach, conf, 2 events, event_reach' (0x8024) Notice that it is almost one hour after wake up when ntp decides to jump the clock back by two whole seconds, instead of adjusting it slowly. Why? Supposedly, the suspend routines should restart ntp, but it is not happening. I have, in /etc/sysconfig/powersave/sleep: SUSPEND2DISK_RESTART_SERVICES="autofs slmodemd irda ntpd" Obviously, the service has not been restarted, but it was left in memory, thus the error on wake up. Why? How to force it to be stopped and restarted with suspend/wakeup? - -- Cheers, Carlos Robinson -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.2 (GNU/Linux) Comment: Made with pgp4pine 1.76 iD8DBQFF0FSytTMYHG2NR9URAo6lAJ0VeeNxZ3UiceCrmK1uNNQbGUT/YgCcCgAx br3nfp3dB2497EeNukj+JPs= =Y6Gl -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Carlos E. R. wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Hi,
I just have seen something in the warn log and I'm investigating. See the detailed log:
Feb 12 12:26:54 nimrodel postfix/local[29204]: ED7FE86F23: to=<cer@localhost.nimrodel.valinor>, relay=local, delay=62, status=sent (delivered to command: /usr/bin/procmail) Feb 12 12:26:54 nimrodel postfix/smtpd[29128]: connect from localhost[127.0.0.1] Feb 12 12:26:52 nimrodel postfix/smtpd[29126]: disconnect from localhost[127.0.0.1] Feb 12 12:26:52 nimrodel amavis[26348]: (26348-06-28) FWD via SMTP: <opensuse-es+bounces-6521-robin.listas...
Notice the clock jumping back two whole seconds? Well, postfix gives notice:
Feb 12 12:26:52 nimrodel postfix/qmgr[4707]: warning: backward time jump detected -- slewing clock ... Feb 12 12:27:34 nimrodel postfix/qmgr[4707]: warning: backward time jump recovered -- back to normality
Now, why did the clock jump back two whole seconds? Is postfix adjusting the clock, or is it just noticing it? I'm using ntp, it shouldn't do that! But it does. See the xntpd log:
12 Feb 03:02:53 ntpd[4635]: offset -0.003808 sec freq 10.337 ppm error 0.010344 poll 10 12 Feb 03:03:28 ntpd[4635]: synchronized to 193.127.101.30, stratum 2
Hi Carlos, I wonder if what I am experiencing is in any way related to your situation. Since installing 10.2 using the now 'default' ext3 file system (in recent past versions always used reiserfs) I have been having problems with either shutting down the system and/or booting 10.2. When shutting down I would almost always be either returned to the login screen (where you enter your username and password) or, the most prevalent, I would end up I believe at level 3 command line prompt (black screen background) and be asked to login as root by providing the password. Here I would login and then run the 'shutdown' command to shut down the computer. At other times 10.2 would either simply attempt the shutdown but then 'lockup' and just keep running with Suse's opening blue splash screen and I would have to press the 'big red button', or 10.2 would actually shutdown normally. Using the available ways of shutdown/restarting made no difference (I recall someone stating that he got the right result when he waited until the screen went into B/W mode before pressing the Confirm (the shutdown) box). At the other end, when starting the system there would be problems as well and very often the only way I could start 10.2 was by booting from the installation DVD because after switching on and going past grub's menu 10.2 simply would not comptete the boot sequence - the splashscreen would stay on with the little dots going round and round the geecko graphic). The other way out of this situation that I took was to push the reset button and this would *sometimes*, and with much time delay, get 10.2 started. I put up with this behaviour as being just an annoyance in 10.2 and would disappear in 10.3 (or if I chose to reinstall 10.2 but this time under reiserfs) until a couple of days ago - and this is where this may be of relevance (at least I hope so because maybe both our troubles would be solved). What I did a couple of days ago was to run e2fsck on the 10.2 partition - and I got the following error messages which I have never seen even when using ext3 some years back: Error message #1: Superblock last mount time is in the future, Fix<y>? Error message #2: /dev/hda9 has gone 49709 days without being checked, check forced [that's 136 years!] After the above was repaired by e2fsck the system shutdown OK. BUT, this morning....... I had the usual trouble of booting. I did the e2fsck check and.... Error message #1 came up again (but not #2). After the fs was repaired 10.2 booted without a hitch. Just to allay what the state of the clock is, the system clock *is* set to the correcttime and date and NTP runs everyday (and I also occasionally run NTPDATE manually just to 'keep NTP honest'). If this is not somehow related to your situation perhaps you (or someone) can offer a reason for Error #1 (and #2), how to avoid it, and what else to check to see what is causing it. Ciao. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Tuesday 2007-02-13 at 11:04 +1100, Basil Chupin wrote:
Carlos E. R. wrote:
That was 2007-02-12... I had a bad "system failure" and I have been busy recovering. I'm now fully on 10.2, and I can no longer check my original problem (was on 10.1).
Hi Carlos,
I wonder if what I am experiencing is in any way related to your situation.
Let's see. ...
What I did a couple of days ago was to run e2fsck on the 10.2 partition - and I got the following error messages which I have never seen even when using ext3 some years back:
Error message #1:
Superblock last mount time is in the future, Fix<y>?
Yes, I have seen that - but here in this list.
Error message #2:
/dev/hda9 has gone 49709 days without being checked, check forced
[that's 136 years!]
Probably a consequence of the previous one.
After the above was repaired by e2fsck the system shutdown OK.
BUT, this morning....... I had the usual trouble of booting.
I did the e2fsck check and.... Error message #1 came up again (but not #2). After the fs was repaired 10.2 booted without a hitch.
Ah, then #2 wouldn't be a consequence of #1.
Just to allay what the state of the clock is, the system clock *is* set to the correcttime and date and NTP runs everyday (and I also occasionally run NTPDATE manually just to 'keep NTP honest').
If this is not somehow related to your situation perhaps you (or someone) can offer a reason for Error #1 (and #2), how to avoid it, and what else to check to see what is causing it.
No, not related. But it was reported here, and the person having it solved by changing the time settings in Yast so that the hardware clock (or cmos, battery clock) uses UTC/GMT time - which will be very inconvenient for you if you double boot to windows, but the preferred method for a Linux only machine. I don't recollect the thread name, but it won't be difficult to locate. It appears that when the fs is checked the system clock is not properly set up yet and is using the bios clock, which if set to "local time" in you case would be 11 hours different from the UTC time that the fs probably uses... and thus the problem. Why this happens in 10.2 and not in 10.1 I can't guess; maybe the time check is new, or it is a new bug. - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.5 (GNU/Linux) Comment: Made with pgp4pine 1.76 iD8DBQFF18OStTMYHG2NR9URAlWpAJwJFk2Dj2CqSna7uRJ3hahkScudVQCcC2rl 3pdaIFAqjeKi7r1ceP2w80I= =bKhi -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Monday 12 February 2007, Carlos E. R. wrote:
Obviously, the service has not been restarted, but it was left in memory, thus the error on wake up. Why? How to force it to be stopped and restarted with suspend/wakeup?
You might be able to force a restart by replicating one of the scripts in /etc/pm/hooks/ Simply having a script there that knows how to do what you do manually seems to do the trick. You can copy one of the existing scripts, change the names to protect the guilty and file it with a leading number to indicate a rough order. I was having problems with sound not working upon resume and this little script seems to have fixed it for me. I suspect something similar might work for you. This one has no shutdown logic, only resume logic. Sample script: Remember to chmod 755 /etc/pm/hooks/##YourScriptName Script starts below #!/bin/bash # # Hack by John Andersen jsa@pen.homeip.net # mostly taken from the powersave project # Probably way more brute force than is needed. ############################################################# # Trigger alsasound upon resume, needed on some intel # high definition audio systems or any other time you find # alsa now working upon resume. # kick-alsa() { /etc/init.d/alsasound restart } case $1 in thaw|resume) kick-alsa ;; esac exit 0 -- _____________________________________ John Andersen
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Monday 2007-02-12 at 18:58 -0900, John Andersen wrote: # Specifies the string to use when sending a message with no to or cc. ^^... new bug to investigate? I dunno where that string comes from... something in Pine.
On Monday 12 February 2007, Carlos E. R. wrote:
Obviously, the service has not been restarted, but it was left in memory, thus the error on wake up. Why? How to force it to be stopped and restarted with suspend/wakeup?
My original problem is now moot. My 10.1 system was almost fully hosed (I blame (80% probability) a faulty 80 wire IDE cable). My root partition (ext3) was badly trashed. Most of my reiserfs and XFS recovered well (manually and with much care), and another ext3 partition was also fine. What matters is that now I'm using 10.2 (upgraded from my previous full backup, which was made of 9.3), so I can not check my original problem. I have new problems, obviously.
You might be able to force a restart by replicating one of the scripts in /etc/pm/hooks/
Simply having a script there that knows how to do what you do manually seems to do the trick. You can copy one of the existing scripts, change the names to protect the guilty and file it with a leading number to indicate a rough order.
In 10.1 there were two settings that usually worked. In the file "/etc/sysconfig/powersave/sleep" I had: POWERSAVE_UNLOAD_MODULES_BEFORE_SUSPEND2DISK="usb_storage sbp2 uhci_hcd stir4200 rt2500 prism54" POWERSAVE_SUSPEND2DISK_RESTART_SERVICES="autofs slmodemd irda xntpd" You see, I did have 'xntpd' there, but it appears it didn't work. Now, in 10.2 -- hold on... I read something in my previous 10.2 test install that I can't find now... Ah: /test_a/etc/sysconfig/powersave/sleep: # DEPRECATED! Configuration for suspend is done in /etc/pm/config # # The settings made in this file are ignored. The new method of doing # suspend is pm-utils. All related code will be removed from powersaved # very soon. For more information have a look at # http://en.opensuse.org/Pm-utils Curiously enough, my upgraded 10.2 install doesn't contain that warning. Another bug, I assume. Ah! I found it. The upgrade process has appended text to the configuration files, so that I have duplicated settings... POWERSAVE_UNLOAD_MODULES_BEFORE_SUSPEND2DISK="usb_storage sbp2 uhci_hcd stir4200 rt2500 prism54" POWERSAVE_UNLOAD_MODULES_BEFORE_SUSPEND2RAM="usb_storage sbp2 uhci_hcd stir4200 rt2500 prism54 POWERSAVE_UNLOAD_MODULES_BEFORE_STANDBY="" ... ... # DEPRECATED! Configuration for suspend is done in /etc/pm/config # # These modules will be unloaded before entering suspend to disk # See README.unload-modules for more information. # "NONE" means don't unload any modules. UNLOAD_MODULES_BEFORE_SUSPEND2DISK="" Yeap. Good job... :-(
I was having problems with sound not working upon resume and this little script seems to have fixed it for me. I suspect something similar might work for you. This one has no shutdown logic, only resume logic.
Sample script: Remember to chmod 755 /etc/pm/hooks/##YourScriptName
Ah, that's the new method. I will have to investigate it. I will also have to find how to disable suspend from altering grub, because it fucks it completely - I use the strong word because it made my system unbootable: stayed waiting after stage 1.5. I used the rescue dvd to boot the installed system (it ignored the suspended system), rewrite grub 18 sectors, reboot, then it tried to recover the suspended system, with a hosed filesystem after the previous boot. :-(
Script starts below #!/bin/bash # # Hack by John Andersen jsa@pen.homeip.net # mostly taken from the powersave project # Probably way more brute force than is needed.
I will save this for later study and maybe use, thanks :-) - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.5 (GNU/Linux) Comment: Made with pgp4pine 1.76 iD8DBQFF18nwtTMYHG2NR9URAlqSAKCC5gbOvIBOzJg1YkRw/vJxmaVgmgCfe4lA 8JlPUrM2/y4RcnFVhOr353g= =/CiN -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Carlos E. R. wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Hi Carlos, There is this big void above because I accidentally deleted your message to me in this thread where I mentioned (13 Feb) that I am getting an error message showing that, in the ext3 partition where the OS [10.2], 'Superblock last mounted in the future'. Your response suggested that this (or similar) was discussed already and that setting the clock to UTC rather than local time was the solution. I have never set the clock to UTC because I do dual-boot but have never had problems with shutting down or booting and have never experienced this error message when everything was formatted in ext3 (and ext2 earlier). You also suggested that this could be a subject of a bug report. Perhaps it is. Could you suggest what its destination could be- suse or ext3 developers or elsewhere. Cheers. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 The Wednesday 2007-02-21 at 00:13 +1100, Basil Chupin wrote: (my mail client thinks you live in tomorrow - not aware of time line, I guess ;-)
There is this big void above because I accidentally deleted your message to me in this thread where I mentioned (13 Feb) that I am getting an error message showing that, in the ext3 partition where the OS [10.2], 'Superblock last mounted in the future'. Your response suggested that this (or similar) was discussed already and that setting the clock to UTC rather than local time was the solution.
If you deleted too much, now you can get the archive in mbox format and import any email into your client :-)
I have never set the clock to UTC because I do dual-boot but have never had problems with shutting down or booting and have never experienced this error message when everything was formatted in ext3 (and ext2 earlier).
I know. Or I guess, rather, or people would have mentioned this problem before. :-) Even so... try changing your time setup, tell Yast to change to some other time zone, and back. Check if related rpms are installed: I'm thinking of timezone or similar name. Just in case it does the trick.
You also suggested that this could be a subject of a bug report. Perhaps it is. Could you suggest what its destination could be- suse or ext3 developers or elsewhere.
I suppose suse/novell bugzilla would be fine: they know better the possible interactions in the system, what do the start up scripts and such. - -- Cheers, Carlos E. R. -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.5 (GNU/Linux) Comment: Made with pgp4pine 1.76 iD8DBQFF2xP8tTMYHG2NR9URAqaIAJ46O1gDAiCy2lT9d/inDwVG3HBd7gCfS4wM RIS1wzGeqQkiwn9wtaIXUVI= =DJgM -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
participants (3)
-
Basil Chupin -
Carlos E. R. -
John Andersen