Jump to content


  • Content count

  • Joined

  • Last visited

Community Reputation

0 Neutral

About derringer

  • Rank
  1. The thing is, all of these options are on the same screen, and it appears that none of them work. Whats even more frustrating is that half of them don't even work on MACs. Things I can't get to work: 1. You cannot 'shutdown a Windows machine after backup completes' 2. You cannot 'sleep a Windows machine after backup completes' 3. You cannot 'wake on LAN' a MAC OSx, it never works, regardless of whether you use Proactive or Scheduled. 4. If you 'wake on LAN' a PC, it gives you 30 seconds for the machine to be responsive, and often times it fails the backup from trying before the client reloads on the PC. None of these are documented. All of them should work... I can hack together DOS batch scripts that perform these functions in about 15 minutes. Why can't EMC get them working? Its mind boggling... these are not huge programming issues... my 15 year old Computer Science degree is sufficient to do them in a few hours time. I am hereby offering my programming services to fix your WAKE on LAN and Sleep code... Wake on LAN is a UDP port 8 issuance and is unbelievably easy. Shutdown and Sleep are easily accessible via Windows APIs. I could seriously implement them all if I had your source code in less than a few days time. But, so that I don't hijaack the original thread poster anymore, his question is: "Can Retrospect Sleep Windows Machines after backups if the checkbox that says it can is checked or not?"
  2. I'd like to clarify a few points: 300MB/sec is not really possible, so we need to figure out exactly what you are talking about. This would be 18000MB/min (which is what Retrospect quotes in its performance statistics,) and that is far from possible for a single workstation backup. I have done extensive testing on extremely high end hardware and I am able to get somewhere around 10000 MB/min with 8 threads running to 6-10000 RPM disks in a high-end RAID0 array and using 4 Gigabit server ports incoming to the server, all close to saturated. But 18000 simply is not possible in Retrospect, and I say that regardless of what hardware you could possibly string together. In fact, this test performance setup is running a Core i970 chip and has 6GB of DDR3-1850 memory with an Areca 1600 series Raid card. With that, I see sustained data transfers (FROM 8 SIMULTANEOUS EXECUTIONS,) of around 200 MB/s with spike of around 450MB/s, but it still doesn't reach what you've quoted. To expound further, my testing shows that a single backup client, even running extremely fast SSD drives on the client end and backing up to the monster server I have described above, cannot crack 45MB/s (2600MB/min,) at its peak on a single large file. Do you mean MB/min, perchance?
  3. To add to this: I cannot get 7.7 to 'shutdown' a windows client after backup either (is there something more than checking the checkbox that one has to do?).. Have tried several XP/32 clients thus far, and I've tried scheduled backups thus far.
  4. derringer

    Wake On LAN problems and questions

    I've done some more testing, so I wanted to report back.. Firstly, the MAC wake on LAN doesn't work in proactive backup either, by my testing. I can send a wake on lan packet from any pc on my network and it wakes the MAC, but Retrospect doesn't appear to ever try. Secondly, on the PC side of things, the Wake on Lan wakes the PC, but doesn't wait long enough for it to boot (times out at around 40 seconds, and the machine takes 60 seconds to boot.) I've also tried it on a hibernate, which comes out in approximately 20 seconds, but Retrospect reports an error as it attempts to backup the client before the client loads all of its software backup, giving a volume not available error.) What I am now doing is using Windows Scheduler with a DOS CommandLine Wake on Lan utility, which I have scheduled in windows scheduler to wake each machine a few minutes before their scheduled backups. This works, but it begs the question why 7.7 advertised "wake on LAN", if it wasn't even tested. I don't know of a circumstance where the Wake On Lan will work regularly, and it doesn't work *at all* for MAC clients, proactive or otherwise.
  5. derringer

    Wake On LAN problems and questions

    I'm not able to check at the moment, but I downloaded the latest MAC version from the website for use with 7.7..
  6. derringer

    Wake On LAN problems and questions

    I'll assume that getting this to work with a non proactive backup script is on the agenda, since there is literally nothing preventing the functionality from working on a time-scheduled backup. But, to answer your question, I did try a proactive backup script, and ran it for about 15 minutes and it never found the client or even attempted to wake it up. It could be I do not understand how proactive backup works, but shouldn't it have tried once in the 15 minute active period to wake up the MAC? It did not.
  7. I have a couple of issues with the new Wake On Lan functionality, and thought I'd bring them up here. Firstly, it appears that 7.7 does *not* wake up MAC clients at all. In fact, from logs, it appears it doesn't even try. I can send Wake On LAN packets from a PC on the network and they wake the MAC up perfectly, so the only thing I can come up with is that Retrospect simply didn't 'turn on' the functionality for MACs. Does anyone have any idea why or when they will be turning this on? The checkbox exists for the MAC client, so why wouldn't we insert the 5 or 6 lines of code necessary to get it to wake the client like the PC clients do? Anyway, please report any experiences anyone else here has had with this. Secondly, waking a PC client works, but Retrospect doesn't wait long enough for my client to boot. Is there any way you can change a setting or preference to get it to wait one minute instead of 30ish seconds, like it currently does? This seems to be something that should be a user adjustable setting. Any help with either of these topics, or even observations, would be much appreciated..
  8. derringer

    Windows 2008 Server x64

    Well, we continue to have the issues with this test server. I experience the *same* issues with a crossover cable directly attached from one computer to the other, with zero other NICs enabled (all other NICs other than that one are disabled.) I have no idea what the problem is, but I'll be saddened if we have to move back to 2003, as when the backup is actually running before it times out, it consitently hits 30+% utilization on a Gbe network link, which is a full 60+% faster than retrospect on a win2003 server. Since the purpose of this hardware upgrade was for backup speed, we really need to get 2008 R2 working. Are all of you who have it working on 2008 R2 x64 ? I'm at a loss as to what the problem is..
  9. derringer

    Windows 2008 Server x64

    I've been unable to get Retrospect to complete a single backup running on Windows Server 2008 R2 yet. I get error -519 after between 1 and 10 minutes of copying, and the 'retry' is never successful after that. I am in the midst of troubleshooting it, but it is not looking good... Anyone successfully running on 2008 Server R2 as the backup server backing up other clients over the network?
  10. derringer

    Lets talk modern hardware performance

    -Firstly, I'm testing backup speed, and there is nothing faster at writing to disk than RAID0. Secondly, this is online backup, which then goes to tape at regular intervals, so I don't need it to be fault tolerant at all. The other server I'm backing up is the old retrospect backup server, so it was set at RAID0 for performance as well. -This would be one slot... I do use 4 slots during normal backup windows, but all of these tests are from 1 slot backups to test single computer backup performance. -I've tested with and without Jumbo Frames. Jumbo frames actually *decrease* performance for small files, and increase performance for large files. The testing on the old server proved that jumbo frames should be turned *off* for retrospect backups of normal machines with lots of small files. If you have a significant number of large files, or even a few very large files, you might want to have it on.. -I'm not using encryption. In these tests I'm also not using compression, as my notes from previous testing showed a decrease in performance without a huge increase in space saving. -The test basically includes a normal O/S install and a small retrospect data directory. I think its probably on the small side for file sizes in comparison to a typical system. Network bandwidth between the two has been measured at between 85% and 95% of gigabit speed, sustained. This is mostly because that is about all the old backup server can sustain, as the new one is faster. Windows File Sharing and DOS window copies yield 6,000 - 7,000 MB/min . This was why I started questioning just what Retrospect is doing that takes it so long to transfer data between these two machines. The actual performance of retrospect is extremely low on high-end hardware is my conclusion. I was hoping someone could post what they see on modern hardware backups to see if there is something I am doing wrong, or if we should investigate another backup solution that would be faster..
  11. Hello all. I wanted to reopen the discussion on retrospect multi server performance. Basically, I want to see some of the statistics others are getting because I've just installed new hardware and was amazed to find that I saw little to no speed improvement. It is my opinion, and I could be wrong, that Retrospect is throttled for some reason and does not scale well to modern server hardware and/or networks. What kind of backup speeds are people on this forum seeing between a fast client and the server? Under the best possible and average conditions, what are your ballpark figures? I am operating on a dual Xeon 5405 server 2Ghz (8 cores,) 8GB of RAM (addressing only 4 because of 32bit) in Windows 2003 SP2 32bit Server with dedicated Areca Raid controllers with battery backup on the PCIe bus employing a Raid-0 striped SATA array. The array is capable of 400MB/sec read/writes tested and confirmed to and from. The server also has 5 NICs all capable of operating in a load-balancing pool, but which I have only tested one individually at a time because of Retrospect's inability to make use of even a single full gigabit pipe. The server also has an Adaptec 39160 SCSI controller running a Quantum LTO-3 tape drive. Now that the server hardware has been explained, here is the client being used for testing purposes, although I have tried several other clients and this is the fastest I havefor testing: Pentium 4 @ 3.4GHZ, 3GB RAM, Windows Server 2003 SP2 32bit, Dedicated Adaptec Hardware RAID controller with 4 RAID-0 striped SATA drives capable of 130MB/sec read/writes tested and confirmed. This client has 2 Gigabit network cards capable of near line speed during testing, although not quite as capable as the server above. What would those of you running multi-server and multi-client retrospect backup networks expect for performance with this kind of setup? Would it surprise any of you to know that I cannot get Retrospect to use more than 26% of a single gigabit NIC at any time during the backup, and that is on the largest of files? I'm getting around 1000MB/min averages for this setup, and frankly, for the capability of the hardware involved, that is just unacceptable. I've even turned off MD5... I would like to ask in this thread that we keep things to citing specific hardware/software examples from each of your networks with ballpark MB/min averages and peaks for backing up typical clients and servers. If I can determine, for instance, that windows 2003 sp2 32-bit is holding back performance and I need win 2008 server or vista 64, I will try that in a heartbeat. I am mostly curious what kind of performance some of you have seen in the real world so that I can tweak our setup and make use of the new hardware we've purchased. I expect that overhead, disaster recovery, cataloging, and the other functions retrospect does will slow down theoretical backup rates, but I would expect that we could at least get to somewhere around 40% of what is possible with a simple windows copy, and I'm nowhere near that...) Thanks for any input anyone can give to give us all some idea of what is possible with the software, before I explore other product options..