Jump to content

Don Lee

  • Content count

  • Joined

  • Last visited

  • Days Won


Everything posted by Don Lee

  1. More data on missing reports. I also see the delay in populating the console, and the reports seem to be the last things to appear. Yes, it sometimes takes a while. I saw some anomalies in the way this was reported, where sometimes the reports were there, and other times they were missing. I run normally with two engines attached to the console. One is the "local" machine (fast ethernet connection) and the other is the "remote" system (connected over the net, slow connection - 25 KBytes/sec?) When I remove the "local" engine from the console, the reports appear reliably for the "remote" system. When I re-add the "local" system, the "remote" reports vanish again. (Usually... It seems to depend on what order the remote engines appear in the sidebar, though. I have seen coredumps (large) while playing with this.)
  2. The console gets confused and crashes in the following scenario. It's obscure. I have 3 Retro installations. One is local, and two are remote, behind NAT. I get access to the remote installs via an ssh tunnel to a local "firewall" machine. Because Retro does not allow me to specify the port number of the connection from Console->engine, I "share" the IP and only connect one remote engine at a time. The local engine stays connected because it is on a local, accessible machine. When I connect the console to a remote machine, I set up the ssh tunnel, and connect Retro to the proxy/tunnel machine. If I kill that tunnel, and start up the tunnel to the other remote machine, the console loses its connection to the first engine, and establishes to the second remote engine. The trouble starts on that second "tunneled" connection. The left sidebar picks up the name of the engine, and puts the little "lock" symbol up. I have to double-click the lock to get the console to request a password. When I enter the password, the local machine loses its "activities", "past backups", etc. It is reduced to a single icon. The second remote machine picks up the new name, but never displays its sub-parts. I soon get a SPOD and the console hangs, being hard to kill. When I restart the console, it works fine.
  3. When you see horrible performance, take a look in Activity Monitor, and see if noifyd and the Retro engine are each eating about 50% of a CPU. If so, kill notifyd and see if the performnce improves. I've seen this several times. It may or may not be related to what you are seeing.
  4. I take it back - I had created a report on my retro console, and went to this forum to post my success. While posting, the other "default" reports magically reappeared. Adding that single report must have "fixed" something.....
  5. For what it's worth, I went rogue, and read the documentation. ;-> Reports can be created from several of the oher panels, with the little "plus" sign in the upper right. It's not blindingly obvious, but quite usable. I can create new reports, but the "default" ones are still gone. I have no custom reports, so I don't need to fix it, only to add the reports I care about.
  6. I have the same problem. A machine upgraded from Retro 8.2 -> 9.0.1 (401) and all the reports are gone. There is no way I see to get them back.
  7. Three times now, I have seen this with Retro 9.0.1 (401) on two different machines. This time I got some debug information. Retro 9.0.1 (401), OS is Mac OS X 10.5.8, Intel mini. (2006, dual core), 2 GB memory. All data on external FW disk. Operation: copy contents of media set from rebuilt 6.1 backup set to a new "disk" media set. No clients involved. Behavior: the operation was proceeding, but appeared to get slower and slower over time. We're talking SLOW (2.9 MB/min) I noticed that one of the CPUs was pegged, with 49% in RetroEngine, and 49% in notifyd. When I killed notifyd, the slowness stopped, and the operation finished. This time, I did a sample of notifyd and the Retro engine, in hopes of telling what it's doing in this tight little loop. (enclosed zip file contains 2 sample files, compressed) Embrace.zip
  8. I had hoped that Retro 9.0.1 (401) would allow me to do off-site backups of critical data with WebDAV, but experience is teaching me that it does not work reliably in 10.5.8. From what little debugging I have done, it looks like the problem is in the WebDAV filesystem in Mac OS X, not in Retro, but the effect is the same. It appears that there are some timeouts in the filesystem, and if files are too large, or networks too slow, the kernel hangs, unkillably, and I have to power cycle the machine to get it back. I remember that I tested Retro 8 with 10.6, and the NAS functions were "much better", so I bet that my real fix is to upgrade the retro server to 10.6, but that may or may not be a good option. In the meantime, beware of WebDAV usage with the retro engine on 10.5.8.
  9. I have seen a variation of this. Client 6.3.029 running on Mac OS X 10.5.8 (PPC mini), newly installed. Server 9.0.1 (401) running on 10.5.8, Intel mini, same network segment - 192.168.0.x The server would not "see" the client, so I could not "add" the client. I added it manually by putting in the IP address, and it added fine.
  10. Welcome back. I like the news. I now have 3 installations of Retrospect and want to report some things that are bugging me. I have long followed a regimen where I run a tape/file backup for some period of time and "rotate" the backup set(s) by manually selecting the media set and using the options->new media option. This allowed me to "rotate" to the new set, but did not force me to update all the scripts. As far as I can tell, this can't be done in Retro 9 - except - as part of a backup operation. Further, that backup operation has to be *scheduled*, because the only place I see to select the "new media" action is where the scripts are scheduled. I also liked that Retro 6 let me tweak the name of the backup set when I did the "rotate". Retro 9 gives me no option. It adds the [00N] unconditionally - and there is no way to rename either the media set or its members. Am I missing something? Can this be done, or do I have to live with the "fixed" names, and running "special" scripts to "rotate" my media sets? Is there some stylistic usage pattern here that I'm missing? Thank you, -dgl-
  11. I finally tried Retro 9.0.1 on a 10.5.8 machine and it still crashes consistently when you try to mount an AFP share from the engine. (Engine running on 10.5.8. If it doesn't crash the first time, try a couple more times. It'll get a kernel panic.) I would think this could be "fixed" by simply disallowing AFP mounts on 10.5.8. I would *much* rather get an error message from retro than a kernel panic. -dgl-
  12. Has anyone out there restored files from 6.x media? What media type? How? My attempts with the beta were all unsuccessful, and I want confirmation that this actually works before I invest more time testing. Thanks
  13. Don Lee

    Retro 8.2 restore from 6.x media

    Thank you for your replies. Has anyone successfully rebuilt a "file" backup set? Most of my data is in that form, and I don't want to transfer it to tape to coax 8.2 into reading it. ;->
  14. I have been using Retro 8.1 (Intel engine machine, no local console usage, 2 clients) at a customer site. I set it up to send me e-mail, and this is very helpful. Comments: 1. It would be really nice to have some sort of option to get fewer e-mails that combine some of the information - like a "digest". 2. The e-mails I do get are not very informative. They could/should contain a little information about the backups when they complete rather than simply when they start, and when they end. 3. It would be *really* nice to be able to periodically e-mail reports out. I know this is probably already on the feature list. -dgl-
  15. I just posted to "bugs" about my experience. My config does not meet Robin's "minimums", but my workload - the GUI - should not overwhelm my platform. Is anyone using PPC successfully? If so, what are you running, and what did you have to do? Thanks, -dgl-
  16. Is there any syntax/technique to specify the port number to the console when setting up a server? The default is port 22024 from what I see here: http://kb.dantz.com/display/2n/articleDirect/index.asp?aid=9705&r=1.008242E-02 Common syntax is to allow an alternate port configuration with: XX.YY.ZZ.QQ:PPPP For instance, - for IP and port 44000. I have tried this in the dialog for adding a server in the console and my input seems to be ignored completely. (the IP is not added, either) This is important when needing to set up SSH tunnels to multiple machines through a gateway. Another compelling reason to provide this ability is to be able to access more than one machine behind NAT. SSH tunnels work well for this, too. This is much harder to deal with if you are forced to use the same port for all machines. Thanks, -dgl-
  17. Don Lee

    Anyone got PPC in production?

    Judging from my experiments with PPC (on an unsupported Mini 1.5 Ghz single CPU), I'm betting that there are some performance problems that are covered up with higher power CPUs. If this were my software, I would be particularly suspicious becuase I am seeing variations in performance that are otherwise unexplained. Sometimes the xfer/verify speeds are good, other times really poor. Are you also seeing wide variations, or is it always poor? -dgl-
  18. Nice -20 is dangerous. Priorities inside the OS are pretty important, and "negative" nice for user processes is restricted to root for a reason. Even nice -15 is aggressive. Running at nice -20 on a busy machine is likely to run into deadlocks and/or starvation under heavy load, or at seemingly random intervals. (i.e. hangs and crashes of the OS) A minor gain of responsiveness at the cost of stability is a bad trade. -dgl-