×

Loading...
Ad by
  • 推荐 OXIO 加拿大高速网络,最低月费仅$40. 使用推荐码 RCR37MB 可获得一个月的免费服务
Ad by
  • 推荐 OXIO 加拿大高速网络,最低月费仅$40. 使用推荐码 RCR37MB 可获得一个月的免费服务

@

file1 is not generated by OS , but a transaction logs produced by a commercial

transaction system built on top of Oracle and BRS. This file is more important than oracle archived/online logs in our backup strategy.

we have daily full backup (snapshot), for any crash occurs in daytime, we can restore our system to 3:xx AM and replay transactions logged in the file1 to bring the system up or close to the crash point.

That's why I want to duplex file1 to file2 on another machine even though it's certified as unreliable, unless
the vendor can enhance their app by allowing log duplex.

I will take a look at Rsync.
Report

Replies, comments and Discussions:

  • 工作学习 / 科技领域杂谈 / How to real time copy /sync/output a file (solaris) to another machine (solaris) ?
    of course i can use 'nohup tail -f file1 >> file2'. However, file2 must reside on the same machine as file1.
    Our security policy does not allow NFS. Therefore, this method does not seem an option to us.
    On the other hand, scp unlike tail -f can provide realtime sync function. The only thing we can do with Scp is to cron a more frequent running schedule, which is really not ideal. Even a five-minute gap can lead to a loss of 100 transactions history.

    Any one can shed some idea?

    Thanks!
    • try rsync
      • A good tool, however
        it seems that this powerful tool can't produce authentic inter-box real-time copy. Thank you.
      • 这个应该是应用解决的问题啊,可以同时有几份log,实时duplex还挺麻烦的,要截取change信息,有些CDP软件有这个功能。
    • why would u do that? if it is mission critical, such remote file is "certified" as unreliable.
      1 alternative is to utilize syslogd to divert all the log message to another log receiving host.
      • file1 is not generated by OS , but a transaction logs produced by a commercial
        transaction system built on top of Oracle and BRS. This file is more important than oracle archived/online logs in our backup strategy.

        we have daily full backup (snapshot), for any crash occurs in daytime, we can restore our system to 3:xx AM and replay transactions logged in the file1 to bring the system up or close to the crash point.

        That's why I want to duplex file1 to file2 on another machine even though it's certified as unreliable, unless
        the vendor can enhance their app by allowing log duplex.

        I will take a look at Rsync.
        • Rsync is no difference in this situation
          It also use ssh to copy file, same as scp. Rync can only benefit when you need to duplicate mutiple files and only some of them are modified, then rsync will only copy the changed file. It's ideal to duplicate folders like website.

          In your case, it's only one file and it always change, so rsync is not helpful at all.
        • so if it is for redundancy , u can just ask ur sysadmin put that log file on one of the file system that has a mirrored backup. so during an emergency exercise, just recover it from the mirrored disk.
          I had thought u were developing a feature. since it is not the case, u can leave it to the UNIX sys admin and let the architect to decide this issue.
          • The disk subsystem is already raid 5+0. Just want to add another layer of data safety
            by copying the file off to another machine in real time mode.
            The application is a commercial one rather than a home-brew one, leaving no place for us to make a change from inside the app.
            • It seems the issue needs to be addressed at architecture level. No matter how frequent the synch is, external file copy can't guarantee "leaving no place for us to make a change from inside the app.".
    • real time is not well defined here. How much delay is allowed? 3 seconds or 0.1 second?
      • pseudo-real time as long as it can match the effect of "tail -f file1"
        Well, we finally decide to tolerate five minutes' data loss. What a shame on us ,hoho...
        Policy has been set up and communicated to our user community. I don't want to be the only paranoid in our department. However, if any Rolian techie guy knows a simple solution for this question and posts it here, that 's great.