On Wed, 02 May 2007 11:21:58 +0200, Bill Maas wrote: >Hi Fubar/RW, > > >A while ago, I wrote: > >""" CF is about a year old, brand "Kingston Technology". """ > >I lied: I just opened up both my soekris boxes and saw that soekris2 >(from which I get the errors) actually has a 512 MB Apacer installed. So >it doesn't seem to be a brand issue. The other box, aptly named >soekris1, has a PEAK CF, but I haven't had any trouble with that one. > >I also didn't mention that I have / mounted ro with an mfs mounted >on /var. For a moment I feared that these were RAM errors from the MFS >partition, but on both error messages that I saved, the sectors were >clearly marked as wd0 sectors. > >I consider soekris2, a net4801, my primary Soekris box, that's why I >gave it the Apacer CF. I try to keep it as "clean" as possible. >soekris1 is a net4521, which serves as a backup for soekris2, and which >I occasionally access for testing. The only writing that has taken place >on the Apacer CF on soekris2 were two OpenBSD installations (3.9 and >4.0), and the occasional remote update, where an update script mounts / >rw, copies a few files, and mounts / ro again. > >You're right about the speed: compared to other brands, the Apacer CF >cards are incredibly fast. But I'm still not convinced about the card's >stability. > Well, we have about a dozen Apacers in the field with OpenBSD installations treating them as if they were IDE drives. All on4801s.
I do the install hdd mode first. Then my concessions to speed and unnecessary writes consist of: vi /etc/fstab :,$s/rw/,softdep,noatime/g ZZ and all the firewalls like that have been running for at least a year with full logging and no problems. The busiest is in a very busy funds management company with branches in three cities and expanding. There is a very busy IPsec tunnel terminating on the firewall and that alone generates loads of logging. Our conservative approach to CF wear outs is to replace the CF at yearly intervals and chuck the old ones into the lab bin MARKED AS NOT FOR FIELD USE. At about $40 AUD for a new one it is not an expensive deal. Now, as to yours below: > >Unusual System Events >> >=-=-=-=-=-=-=-=-=-=-= >> >Apr 15 01:00:39 soekris2 /bsd: wd0a: id not found reading fsbn 367868 >> >of 367868-367871 (wd0 bn 367932; cn 365 tn 0 sn 12), retrying >> >Apr 15 01:00:40 soekris2 /bsd: wd0: soft error (corrected) That CF has something on it ahead of that kernel, 364 cylinders of it in fact. Maybe it was worn by some logging when another OS lived in lower space? I don't know its history but on the CF I use I would never expect an error on /bsd. It lives on wd0a at the start of the "disk" and might just be written over twice in its service life. <snip> With 4.1 patching a few nits that happened between RTM and official release, I'll be replacing a few in the near future. I might just pound the crap out of some with a CF-to-IDE adaptor I have, using idle cycles in a server and see if I can finally kill one. <BEG> As I said b4, Go well! _Rod____ Depressed? Me? Don't make me laugh! :Spike Milligan:1918-2002: _______________________________________________ Soekris-tech mailing list [email protected] http://lists.soekris.com/mailman/listinfo/soekris-tech
