SD-Card defects - FS corrupt

[rfk]

As in several threads one can read about defect SD-cards and corrupted filesystems:


has anyone ever tried to get aufs working with sunxi-linux on the A20 board?
Using FUSE works well for /var but not for /etc (which would be necessary to mount / as read-only)

I did it on the Raspberry Pi - it seems quite a life saver. Even more, as I go towards 24*7


Regards
Richard

[hnb2907]

Hi,

Isidoro and myself have compiled kernel support for F2FS.

I've got a 4GB card (with approx 800mb free), which has been permantly running debian for a month without problems.

A couple of posts that might be of interest to you:
https://www.olimex.com/forum/index.php?topic=2305.0
https://www.olimex.com/forum/index.php?topic=2394.0


Chris.

[rfk]

Chris ,


thanks for the pointer!.
Just a question: as I mentioned in the post here: https://www.olimex.com/forum/index.php?topic=2394.0, I have from Dec. 22. to now approx. 650 Mio writes to /var and /etc (*) - do you have experience with such number of writes and F2FS?


I have /var mounted with FUSE, thus all writes to /var go into RAM-Disk.
Additionaly I fiddled around with /etc/rc*.d/ , so the writes to root-fs during startup are now down to about 300 - which I can live with for now.
Basically all services are started from rc.local after restoring volatile data to /var. During shutdown, volatile data are written to SD. If there is a power outage, these data are lost - however, not an issue, since the whole stuff is supposed to run from UPS - thus "should never happen ...".

My question about aufs is mainly related to writing into /etc - which still could become an issue, putting the box out into the field.


Richard


(*): on a HD of course! This is an application with Nagios and pnp4nagios, writing to RRD's. I still hope, it can be done without adding an external disk - just using RAM-disk and 'some' scripts.