196 replies to this topic

[domi-sat]

also ich hab ne 1,5 TB mit EXT3 und auch bei mehreren aufnahmen nur am anfang diesen freezer.

[darmot]

Zusammengefasst

betroffen sind (mit Ausnahme der glücklichen Benutzer):

-grosse Platten >1TB (bei mir eine 2TB und eine 1,5TB)
-ext4 formiert egal ob per Telnet oder per GUI
-Kernel = 3.x
-Anzahl der Hänger steigt mit dem Befüllungsgrad der Platte; bei mir fing es ab ca. 20% mit den ersten Hängern mitten in den Aufnahmen an)
-tritt auch bei Aufnahmen ohne weitere Nutzung/Last (also Timer in Abwesenheit) des Receivers auf
-Lt. TS_Doctor sind die Aufnahmen fehlerhaft; also kein Fehler bei der Wiedergabe

Getestet aber leider ohne Erfolg:
-Recording * auf diverse und auf hier empfohlene Werte => mehr/längere Hänger
-EXT4 Journaling ausgeschaltet => kaum spürbare Verbesserung

potentieller Workarround (laut diversen Berichten; ich selbst stelle in den nächsten Tagen um):
-interne Platte ext2 formatieren und demensprechen mit UUID mounten lassen

Fakten:
-ältere OpenPLI mit Kernel < 3.x: keine oder höchstens mal einen Hänger am Anfang einer Aufnahme
-keine Hänger bei Aufnahmen auf NAS über LAN
-es gibt viele Benutzer die sich so mit dem Receiver beschäftigen, nicht in Foren aktiv sind und die Hänger als Empfangsstörung deuten/hinnehmen

[MiLo]

what if you disable the "ordered" journal writing? These may cause long hangups in write calls, so might explain your hickups.

To do this, install tune2fs with opkg (you probably have it already) and then, assuming your harddisk is sda1:

umount /dev/sda1
tune2fs -o journal_data_writeback /dev/sda1
mount /dev/sda1 /media/hdd

If you type "mount" now, you should see "data=writeback" in the options for the harddisk mount.

[anudanan]

Do you know from yourself or from your colleagues about problems with ext4 and ordered yournal writing?
some users with ext4 or ext3 with ext4 handling in kernel 3.x have problems with their movie records. There is a long thread inside the et-view-support forum.

Edited by anudanan, 23 January 2012 - 22:34.

[hulster]

Nur Platten >1TB betroffen ist NICHT korrekt. Im Xtrend Forum haben sich mehrere gemeldet, deren Platten < 1TB sind, inkl. mir. Der Thread dort ist mittlerweile über 70 Seiten lang. Es handelt sich also durchaus um ein signifikantes Problem.

[pieterg]

Have you tried with the file /etc/sysctl.conf removed?
the vm values in there have been customized for a 2.6.18 kernel, on a device with limited memory.
The kernel defaults have changed over time, and I can imagine ext4 on a 3.1 kernel would be optimized for default vm values.

[anudanan]

Have you an information, who the default values are different from the 2.6.18 kernel values?

My /etc/sysctl.conf has the following values on the 2.6.18 kernel
vm.dirty_writeback_centisecs = 300
vm.dirty_background_ratio = 1
vm.dirty_ratio = 60

[MiLo]

Those are the settings used on all boxes. Pieter's suggestion was to remove the file and reboot, so that you use the kernel's built in defaults.

On boxes with 512 MB, those settings won't have much effect at all, it's very unlikely that your dirty_pages grows to 200MB in just 5 seconds.

[darmot]

After i removed the /etc/sysctl.conf, the box produces only at the beginning of a hd record a very very small hickup. The hd record and a parallel made sd record looks fine. I already disabled the journaling by tune2fs -O ^has_journal a couple of days before, but not set the writeback mode yet.
This will be done today evening.

[pieterg]

Please test one thing at a time, so we know whether it was the journaling, or the sysctl.conf settings

[familievierfuss]

Please look also to this thread. It describe the same problem.

Edited by familievierfuss, 26 January 2012 - 12:14.

[Hump]

I'm testing ext4 "journaling off - writeback enabled" for quite a while and the freezer are at a absolut minimum. If at all, they only appaer while deleting huge files when recording is aktive.
In addition I have deleted the sysctl.conf as suggested by MiLo. - Thanks MiLo.

[Kosh]

If this is a problem when also deleting files, you should modify the delete speed to say 50 MB/s, this circumvents large hickups while recording.

[MiLo]

In addition I have deleted the sysctl.conf as suggested by MiLo.

That wasn't my suggestion... I'm still assuming that the sysctl settings have no impact whatsoever.

[MiLo]

I'm testing ext4 "journaling off - writeback enabled" for quite a while and the freezer are at a absolut minimum. If at all, they only appaer while deleting huge files when recording is aktive.

Is that a "real" ext4 partition, or just an ext2/ext3 that happens to be mounted by the ext4 driver?

Deleting a 100GB file from the 160GB disk while recording did not cause any problems on my system, where i set the demux size to its minimum to maximize the chances of overrun. The "rm" command took almost a minute to complete.

So it's still a riddle why those bigger faster terabyte drives get such weird latencies on the ET, while an old laptop drive has no problem whatever I do to it.

[darmot]

Please test one thing at a time, so we know whether it was the journaling, or the sysctl.conf settings

But i have already done so.

1. Changed recording data* values => large values causes large freezers
2. Setted Journaling off by tune2fs -O ^has_journal => still freezers
3. deleted sysctl.conf => only a very small freeze at the beginning of a hd record

[Hump]

That wasn't my suggestion... I'm still assuming that the sysctl settings have no impact whatsoever.

You are right and sorry for the mistake. It sure was not your suggestion. Could be, that it has no impact, but I guess, it will not hurt either.

Edited by Hump, 26 January 2012 - 13:33.

[Hump]

Is that a "real" ext4 partition, or just an ext2/ext3 that happens to be mounted by the ext4 driver?

It's a real ext4 partition.

Deleting a 100GB file from the 160GB disk while recording did not cause any problems on my system, where i set the demux size to its minimum to maximize the chances of overrun. The "rm" command took almost a minute to complete.

So it's still a riddle why those bigger faster terabyte drives get such weird latencies on the ET, while an old laptop drive has no problem whatever I do to it.

Could it have something to do with the 4k Sector Drives? My drive is formated with starting sector 2048.

[pieterg]

That wasn't my suggestion... I'm still assuming that the sysctl settings have no impact whatsoever.

You are right and sorry for the mistake. It sure was not your suggestion. Could be, that it has no impact, but I guess, it will not hurt either.

No, it won't hurt, but I think we'd like rule them out.
Could you please re-install your sysctl.conf file, by doing the following:

opkg install kernel-params -force-reinstall

And reboot. If you have no increase or decrease in recording 'performance', we can forget about the sysctl settings.

[Hump]

No, it won't hurt, but I think we'd like rule them out.
Could you please re-install your sysctl.conf file, by doing the following:

opkg install kernel-params -force-reinstall

And reboot. If you have no increase or decrease in recording 'performance', we can forget about the sysctl settings.

I tested several HD recordings after deletion of sysctl.conf and could not find any differences in performance. Still no freezers with ext4-journaling off and writeback enabled.