pdflush and syslogd 100% CPU usage

[geforce]

Any other ideas on what I could do about this? Or is a faster SD card the only option?

Cheers.

[birdman]

Do you know what is doing the writing, and at what rate?

[geforce]

Not really. Top says that it's the [pdflush] process taking 100% CPU, but what is causing it to do so I cannot say.

I've used linux a little in the past, but I wouldn't call myself an expert. Is there any way of finding out what's causing pdflush to do this?

Cheers

[MarkF]

One thing that can make writes to SSD partitions slower is if the filesystem is not aligned to an erase block boundary.  Though you are using an SD card, I would suggest the same thinking might apply.  Or it might not.

Take a look at this link for a bit of information on how this can be done.

[birdman]

Not really. Top says that it's the [pdflush] process taking 100% CPU, but what is causing it to do so I cannot say.

What are you running that is likely to be writing to the filesystem a lot?  What non-standard (i.e. added by you) services are running?

[dattaway]

What kernel version are you using?  I remember reading a lot of diffs between versions over storage on SD devices, on crude things like spinlocks and such.  It may be a bug that has been patched already.

[geforce]

I'll have a look at the erase block boundary, thanks very much.

The only thing that will write to the fs much that I've added is Samba - a directory on the SD card is shared on the network as a shared server space.

I'm running kernel version 2.6.22.18. Is there any way of checking if there are bugs in this kernel version? If so, how easy is it to switch to a newer (bug free) kernel?

[naugtur]

Hi.

I happen to have the same problem in a bit different situation.

When using mysql I installed on the plug (and randomly for no reason from time to time) I get the pdflush 100% and jffs2_gcd_mtd1 after that. When it's with mysql, the mysql gets 100% CPU before all that. But I'm not sure it's mysql causing it, it's rather hanging on a disk access wait.
I use a basic debian that was there and a pendrive (not that slow) as a storage.  I moved all data and logs that I was able to find to that storage.

Whtat worries me is that it might be a nand used-up block that it trips over.

And does the default debian install swap anything to disk?

[rahul]

I had the exact same problem show up on my sheeva this past weekend. Loadavg shooting up to 6-7 and pdflush pegging the CPU. I had made no changes in the software or config so it was most likely a hardware issue. I was still running off the default stock Ubuntu image from the NAND flash that came with the sheeva. When i checked 'dmesg' i saw four messages complaining about 'bad ersaseblock' on the NAND flash and the bad eraseblock addresses reported were in the root filesystem (/dev/mtdblock1) partition. Suspecting that was the problem i copied over the root filesystem image to a 2 GB SDcard and changed bootparams  so that root is mounted from the SDcard. This link was immensely helpful. http://plugcomputer.org/plugwiki/index.php/SD_Card_As_Root_File_System. That solved it! I am no longer using internal NAND flash for my root filesystem and loadavg is back to < 1 with the sheeva responding normally while running all my applications (asterisk, openvpn etc). This link is also very helpful if you want to boot off USB or SDcard  (i.e. have your kernel read from those devices) and have your root filesystem as well on USB/sdcard http://computingplugs.com/index.php/Booting_entirely_off_an_external_USB_device . I guess i should go all the way and also boot off the sdcard instead of just having root on it but for now its working great and the original problem is gone.

[pervez khan]

Thank you, I like it

Look into what is producing a load-average of > 20?