SFS cant find key errors?

[runequester]

I am suddenly getting errors when booting up workbench "find extentbnode cant find key" and various random numbers.

The HD is using SFS, and the error screen specifies its a smart file system error.

What gives and any suggestions on what to do ?

[itix]

Old SFS versions may randomly trash its internal data structures. Try to find newer version and repair your existing partition or reformat.

[wawrzon]

from my experience sfs will always trash the partition its installed on sooner or later, starting with errors. there is no new enough version to be secure. i am currently using sfs on my aros68k partition, but only because i wanted to test it too. well, i have a lot of data traffic on it, updating the whole system virtually every other day to the latest nightly, it is already giving me customary errors almost from the start. but since this is a test partition anyway, im not that concerned, will sure reformat to pfs3 or something as soon as im in mood to copy the data there and back.

[runequester]

So you'd recommend PFS 3 then?
 
I guess Im surprised, since I /think/ two of the partitions on the CF card I received from Amigakit are SFS too, and they've been rock solid.
 
This was a new card that I formatted with SFS, so it's possible there's a card issue, but if PFS is more stable, I'll consider that.

[wawrzon]

did you rewrite data to the cards a lot? seems power usage triggers that behavior.´on the other hand ive been suggested that the certain combination of os elements is responsible for sfs reliability problems. funny enough none wanted to tell me anything specific, which i considered a little a-social. nevertheless its my experience, others may differ.

[runequester]

did you rewrite data to the cards a lot? seems power usage triggers that behavior.´on the other hand ive been suggested that the certain combination of os elements is responsible for sfs reliability problems. funny enough none wanted to tell me anything specific, which i considered a little a-social. nevertheless its my experience, others may differ.

The card is brand new, but I did copy about 2 GB of data to it, over 3 days of time.

[wawrzon]

thats too early to tell. the problem occurs usually for me when you extensively delete and write or replace data on your hd. perhaps its a fragmentation problem. just copying few gig data to it over few days should be almost safe.

actually i consider now to leave sfs on my aros68k partition to observe what happens to it. it gives me errors now and then, but effectively it doesnt seems to corrupt further. perhaps because i overwrite corrupted files with every new nightly.. to be honest im a little too rude with that drive, it isnt a very reliable reference.

[itix]

I am sure your CF card is fine. Old 68k SFS has a bug where it trashes its journaling system when writing to multiple files simulatenously. It is extremely rare but it does happen. Luckily you dont lose any data but just have unwritable partition and certainly it is less hassle than FFS.

It has happened to me on my A1200, and it happened to me on my Pegasos until SFS was fixed.

[wawrzon]

@itix: im not sure if this is the same as the problems i have experienced, have you ever used strohmayers versions?

[itix]

@itix: im not sure if this is the same as the problems i have experienced, have you ever used strohmayers versions?

Hell, no.

[wawrzon]

@itix
well then, i guess you have missed some fun. just look over at aw.net, currently two guys independently have lost theis sfs partitions under os4 yesterday. i dont mention the third who cant see his hd anymore, because its likely a different issue. threads about sfs crashes appear quite on regular basis while i didnt observe similar about the other file systems. in comparison i would say, except for annoying validation orgies, ffs is quite dependable. at least i cant recall any issues.

[itix]

@itix
well then, i guess you have missed some fun. just look over at aw.net, currently two guys independently have lost theis sfs partitions under os4 yesterday. i dont mention the third who cant see his hd anymore, because its likely a different issue. threads about sfs crashes appear quite on regular basis while i didnt observe similar about the other file systems. in comparison i would say, except for annoying validation orgies, ffs is quite dependable. at least i cant recall any issues.

There is one: it is very slow when scanning large directories and files.

[wawrzon]

@itix, you mean ffs? well right, it isnt fast of course, ive never meant that. but its reliable even though revalidation sounds scary and keeps the machine from booting, which might lead to premature conclusion that the drive is broken. first time it happened to me i restarted the machine multiple times while the process because i couldnt understand whats going on. but it easily survives that.

i can also remember thomas richter told us on natami forum that ffs was actually smarter than sfs. dont remember why though, something like multithreaded or the like.

[itix]

@itix, you mean ffs? well right, it isnt fast of course, ive never meant that. but its reliable even though revalidation sounds scary and keeps the machine from booting, which might lead to premature conclusion that the drive is broken. first time it happened to me i restarted the machine multiple times while the process because i couldnt understand whats going on. but it easily survives that.

It is just that validating my 2GB partition on my Amiga 1200 (040, internal IDE) took forever to complete. I cant wait 8 hours to use my Amiga again.

i can also remember thomas richter told us on natami forum that ffs was actually smarter than sfs. dont remember why though, something like multithreaded or the like.

Not necessarily "smarter" but as I have understood FFS is better at handling simultaneous read/write requests (not using by threads but some internal mechanism). Real time applications are not starving due to other application getting all attention from the filesystem.

[zipper]

It is just that validating my 2GB partition on my Amiga 1200 (040, internal IDE) took forever to complete. I cant wait 8 hours to use my Amiga again.

Phew, 1GB on CSPPC SCSI took perhaps one minute...