home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!noc.near.net!transfer.stratus.com!sw.stratus.com!dswartz
- From: dswartz@sw.stratus.com (Dan Swartzendruber)
- Newsgroups: comp.sys.stratus
- Subject: Re: loss of end-of-file mark problem
- Date: 8 Jan 1993 17:34:56 GMT
- Organization: Stratus Computer, Inc.
- Lines: 32
- Distribution: world
- Message-ID: <1ike00INNheo@transfer.stratus.com>
- References: <1993Jan7.131833.15374@porthos.cc.bellcore.com> <1ikbuqINNrv9@transfer.stratus.com>
- NNTP-Posting-Host: redondo.sw.stratus.com
- Keywords: tps
-
- In article <1ikbuqINNrv9@transfer.stratus.com>, moser@sw.stratus.com (Tom Moser) writes:
- > In article <1993Jan7.131833.15374@porthos.cc.bellcore.com>, dante@decoy.uucp (25439-dante) writes:
- > > In the event of a VOS crash, I have heard that it is possible that VOS
- > > can lose a files's end of file mark, thereby corrupting the file and losing
- > > access to the file on reboot.
- >
- > This is not very likely if the file and directory data are already written to
- > disk at the time of the crash. File data and directory information which are
- > updated in the file cache managers memory but not written to disk may be lost
- > in the event of a crash but the data can be flushed to disk using the s$control
- > system call with the runout_opcode.
-
- The problem is that doing a runout after every such write is a terrible performance hit.
-
- [discussion of using TP deleted]
-
- TP is somewhat of overkill in many of these cases. Probably log-protected files
- would be a good solution. LPF basically guarantees that any system call which
- changes file information will be done atomically. I.e. if a record spans disk
- blocks, both blocks will be modified or neither will. Consistency of indexes
- and eof markers is also guaranteed. And unlike TP, *no* application code needs
- to be changed. I believe LPF has been committed for Release 12 (check with your
- SE to be sure though).
-
-
- > -Tom
-
- --
-
- #include <std_disclaimer.h>
-
- Dan S.
-