Alix 2d13 led lights

[MaxPF]

I notice that at some point in the past month with the 2.1.1 builds the led lights started having a "knight rider" effect instead of the usual 1st led on other 2 off. I even tried installing blinkled, but the 1st led is still blinking.

Is this change intended?

[jimp]

Since 2.0 they have done that effect at bootup until the boot sequence is complete. If the lights are still doing that, check the console and see what is being displayed there at the end of the boot sequence.

[jimp]

For good measure, I loaded a current 2.1.1 snapshot on my ALIX and the LEDs are normal at the end of the boot sequence. There must be an error waiting for you on the console output that shows it not getting all the way to the end of the process successfully.

[MaxPF]

Thanks for checking. I will check the console when I get home.

Update:
No errors. I rebooted from console and it didn't happen again, so it's all good now.

[senser]

I get that when I have a fuse popping, causing my alix to not sutdown properly. The next boot the LEDs will continue playing KID. I haven't checked the console either and a "nice" reboot will always resolve this.

[pvoigt]

I am testing 2.1.1-PRERELEASE on an ALIX.2D13. And I am observing this not ending "knight rider" effect of the LEDs over various snapshot. The only issue in the console log is:

Generating RRD graphs...Killed

After various tests and reboots I can summarize as follows:

On every cold boot the RDD error appears which is accompanied by the left-right-left LED flashing ("knight rider effect").

After the first and all following warm boots the RDD error disappears and the LED flashing stops as soon as the boot sequence is finished. The corresponding RDD log is:

Generating RRD graphs...done.

Regards,
Peter

[senser]

Hmm, interesting Peter. One thing I have noticed is that some RRD graph data went void after the sudden shutdown caused by fuse popping, even though I have set pfSense to backup RRD data every 8h. This is on pfSense 2.1 with some git fixes.

[phil.davis]

Generating RRD graphs...Killed

It ran out of real memory during the boot sequence. Some process, that was the unlucky one to request some more memory when none was left, gets killed. The boot sequence does not finish correctly, so the LED normalize never gets called and thus the lights are left in the flashing sequence.
These days, do anything moderately complex on an Alix with 256MB and this happens.
Usually the system comes up OK, because after the boot some run-time interface event/s happen and that causes OpenVPN restart, Dynamic DNS re-check… so if some of those things were killed during the boot, they get kicked again after boot.
It only really bites if "pf" or "inetd" or some other really basic process is the one that is unlucky and gets killed.

[pvoigt]

@phil.davis:

Generating RRD graphs...Killed

It ran out of real memory during the boot sequence. Some process, that was the unlucky one to request some more memory when none was left, gets killed. The boot sequence does not finish correctly, so the LED normalize never gets called and thus the lights are left in the flashing sequence.
These days, do anything moderately complex on an Alix with 256MB and this happens.
Usually the system comes up OK, because after the boot some run-time interface event/s happen and that causes OpenVPN restart, Dynamic DNS re-check… so if some of those things were killed during the boot, they get kicked again after boot.
It only really bites if "pf" or "inetd" or some other really basic process is the one that is unlucky and gets killed.

Hm, I am not sure, if I get you right: Do you really mean, the RDD error ist memory related? But why does it occur only sometimes - and according to my observation - mostly on cold boot? When writing these lines I've just booted my Alix for an installation of the latest snapshot. This time the cold boot does not produce the RDD error - and the LED flashing stops as expected. That's strange but the relation RDD error <-> Flashing LEDs after the end of boot sequence is very reproducible.

I certainly agree with you that the Alix boards nowadays are hardware limited. But they should still be sufficient for most home users. Maybe I got you wrong but I would NOT expect an Alix board to run out of real memory during the boot process. I would rather expect that under heavy load.

My conclusion is therefore that the RDD code might stil suffer from an undetected bug.

Regards,
Peter

[phil.davis]

It tends to be very timing-dependent. For example, I have 2 OpenVPN site-to-site links, so some time after OpenVPN is started, those links actually establish, and in reaction to that various "link up" event/s are notified to check_reload_status and it does some stuff…
If that happens to happen when RRD graphs are reloading, then there happen to be a bunch of things happening at once. The things that are running PHP scripts grab quite a lot of memory. Something gets "Killed" (not necessarily the RRD graph loading).
The "Killed" is likely often associated with the reloading RRD graphs step, because reloading RRD graphs has a bit of processing to do, and I expect a bit of memory to use while doing it.
For me, I get this issue when I have more than 2 OpenVPN links.

[Xon]

I've noticed the Blinkled package has a memory leak in it.

With 2 LEDs bound to separate interfaces, I've noticed that each BlinkLed process will start at 1mb of memory and climb upto 10-20mb or so. I've setup a cron task to restart the BlinkLed package every 6 hours prevents it getting to horrible.