Re: [9fans] Fossil robustness
Both fossil and kfs seem like the wrong tool for your job. In addition to the robustness questions, they (especially fossil) include features you're not going to get anything out of in your environment. If you use something like paqfs(4) or sacfs(4) (not sure which is more appropriate) you'll get safer operation and have a smaller system as well. PGP.sig Description: This is a digitally signed message part
Re: [9fans] Fossil robustness
erik quanstrom wrote: I've made a customized install procedure copying from the standard one. Where can i find a procedure for kfs to copy from? Can kfs be configured to be absolutely insentive to hard power down ? "absolutely" is too strong. if one turns off atime with kfscmd/atime, it is pretty robust. see mkfs(8) for some tools. but really, just get a ups. you'll be much happier. Oh, yes. It's true, but i can't and, in any case I cannot control the operators :-( - erik
Re: [9fans] Fossil robustness
On Tue Apr 20 06:46:54 EDT 2010, a.vera...@tecmav.com wrote: > Hi all. > > I'm building an industrial application hosted by several independent > cpu server, each of them booted from a CFlash on sdD0. > > The application doesn't write on sdD0 and there are no redirection on > local files in the cpurc scripts. > > In this particular situation fossil should be actually used read only so > allowing to use write protected CF and/or to suddenly power off the system > without > damaging the file system. > > Instead, fossil writes on sdD0 (doesn't boot fom a write protected CF) > and the power loss destroy the file system more than fifty-fifty. > > Unfortunately I cannot guarantee stable/correct operating conditions. paqfs(4) seems ideal for this situation. - erik
Re: [9fans] Fossil robustness
> I've made a customized install procedure copying from the standard one. > Where can i find > a procedure for kfs to copy from? Can kfs be configured to be > absolutely insentive to hard power down ? "absolutely" is too strong. if one turns off atime with kfscmd/atime, it is pretty robust. see mkfs(8) for some tools. but really, just get a ups. you'll be much happier. - erik
Re: [9fans] Fossil robustness
Yes, i've had a lot of problems with fossil when it gets killed. My issue was with wikifs that had some sort of memory leak i suspect, it would fill up the memory, and then fossil would crash and/or get corrupted. I had an idea for a project to use mycroftiv's rootless kernel images and have a script check whether fossil died or what, and if it did reformat it using latest venti snapshot and reopen the root, but i don't know how involved that would be. I would think -r doesn't modify the filesystem if snapshotting is turned off, but i am probably the wrong person to be asked...I currently don't even have my plan9 system installed. HTH On Tuesday 20 April 2010 13:39:59 Adriano Verardo wrote: > John Soros wrote: > > Hello Adriano, > > Have you disabled all snapshotting features? Usiong open -r? > > How are you starting fossil, what's your configuration? > > Hi, John > > fsys main open -AWVP -c 3000 > srv fossil > srv -p fscons > > on /dev/sdD0/fossil > > open -r guarantees that fossil doesn't do physycal write at all or > prevent only user to w/create files ? > > After a fatal power down fossil complains about "metadata corruption" or > "lost 386/init" or > or the corruption of some very first logical sectors. > > adriano
Re: [9fans] Fossil robustness
John Soros wrote: Hello Adriano, Have you disabled all snapshotting features? Usiong open -r? How are you starting fossil, what's your configuration? Hi, John fsys main open -AWVP -c 3000 srv fossil srv -p fscons on /dev/sdD0/fossil open -r guarantees that fossil doesn't do physycal write at all or prevent only user to w/create files ? After a fatal power down fossil complains about "metadata corruption" or "lost 386/init" or or the corruption of some very first logical sectors. adriano
Re: [9fans] Fossil robustness
maht wrote: On 20/04/2010 11:45, Adriano Verardo wrote: Hi all. I'm building an industrial application hosted by several independent cpu server, each of them booted from a CFlash on sdD0. The application doesn't write on sdD0 and there are no redirection on local files in the cpurc scripts. In this particular situation fossil should be actually used read only so allowing to use write protected CF and/or to suddenly power off the system without damaging the file system. Instead, fossil writes on sdD0 (doesn't boot fom a write protected CF) and the power loss destroy the file system more than fifty-fifty. Unfortunately I cannot guarantee stable/correct operating conditions. Any suggestion ? Thanks in advance adriano Fossil is essentially a cache for Venti, the non-Venti special case is what you are attempting to use. Perhaps kfs / cws would be a better choice. I've made a customized install procedure copying from the standard one. Where can i find a procedure for kfs to copy from? Can kfs be configured to be absolutely insentive to hard power down ? Or even booting from a CD. I've only a CF slot. And, in any case, the environment doesn't allow to use mechanical devices.
Re: [9fans] Fossil robustness
Hello Adriano, Have you disabled all snapshotting features? Usiong open -r? How are you starting fossil, what's your configuration? -- John Soros On Tuesday 20 April 2010 12:45:09 Adriano Verardo wrote: > Hi all. > > I'm building an industrial application hosted by several independent > cpu server, > each of them booted from a CFlash on sdD0. > > The application doesn't write on sdD0 and there are no redirection on > local files > in the cpurc scripts. > > In this particular situation fossil should be actually used read only so > allowing > to use write protected CF and/or to suddenly power off the system without > damaging the file system. > > Instead, fossil writes on sdD0 (doesn't boot fom a write protected CF) > and the > power loss destroy the file system more than fifty-fifty. > > Unfortunately I cannot guarantee stable/correct operating conditions. > > Any suggestion ? > > Thanks in advance > > adriano
Re: [9fans] Fossil robustness
On 20/04/2010 11:45, Adriano Verardo wrote: Hi all. I'm building an industrial application hosted by several independent cpu server, each of them booted from a CFlash on sdD0. The application doesn't write on sdD0 and there are no redirection on local files in the cpurc scripts. In this particular situation fossil should be actually used read only so allowing to use write protected CF and/or to suddenly power off the system without damaging the file system. Instead, fossil writes on sdD0 (doesn't boot fom a write protected CF) and the power loss destroy the file system more than fifty-fifty. Unfortunately I cannot guarantee stable/correct operating conditions. Any suggestion ? Thanks in advance adriano Fossil is essentially a cache for Venti, the non-Venti special case is what you are attempting to use. Perhaps kfs / cws would be a better choice. Or even booting from a CD.
[9fans] Fossil robustness
Hi all. I'm building an industrial application hosted by several independent cpu server, each of them booted from a CFlash on sdD0. The application doesn't write on sdD0 and there are no redirection on local files in the cpurc scripts. In this particular situation fossil should be actually used read only so allowing to use write protected CF and/or to suddenly power off the system without damaging the file system. Instead, fossil writes on sdD0 (doesn't boot fom a write protected CF) and the power loss destroy the file system more than fifty-fifty. Unfortunately I cannot guarantee stable/correct operating conditions. Any suggestion ? Thanks in advance adriano