johnkgibson Posted January 8, 2007 Report Share Posted January 8, 2007 Hi all, I am in the process of installing a new OS X server on a G5 Dual 2.3 with PCIe slots... I have the Atto UL5D Adapter, Retrospect 6.1 update.... However when it scans for devices. It freezes. This is what I will be doing tomorrow: Update ATTO Drivers, Firmware and other stuff Update Retrospect Device Drivers And then see if this works. Next is to try a new Drive, Cable, and Terminator... Any other suggestions for me to try on this system? Thanks John Gibson Quote Link to comment Share on other sites More sharing options...
rhwalker Posted January 8, 2007 Report Share Posted January 8, 2007 John, It would be helpful to know your version of Retrospect 6.1 and the RDU. Reason - the help we give in these forums persists long past the solution (hopefully) for your problem, and anyone searching the forums months from now will not get any information from you simply saying you have the "latest" version. Please provide that information for the benefit of others. That said, we had a similar situation on our Xserve G5 with Exabyte VXA-2 1x10 1u (SCSI) PacketLoader attached to our ATTO UL4D. I assume that you have a high-quality LVD SCSI cable and terminator. We found that Retrospect would hang exactly as you describe, while scanning for devices, intermittently. Persisted across various Retrospect 6.1.x versions and across Mac OS X Server 10.4.2 through about 10.4.6 or so. The problem was solved only when we made the Exabyte drive and its autoloader the only devices on that SCSI channel of the UL4D. Previously we had a SoftRAID mirror primary (Seagate Cheetah 75 GB SCSI 10000 RPM) on the same SCSI channel; moving it to the other channel made things work reliably, never a hang since. The Xserve would hang in some sort of a spin-wait loop at high priority, could still ping the server but it never made it down to lower priorities to run any user processes so, for all practical purposes, it was hung. It's my belief that Retrospect somehow caused a SCSI bus reset during its device scan, which caused an interrupt for a data transfer from/to the SoftRAID RAID 1 mirror to be dropped. Because that was our OS volume (and also our VM paging store), the system hung on an I/O transfer that would never complete. Try moving the drive to its own channel, see if that helps. Russ Quote Link to comment Share on other sites More sharing options...
johnkgibson Posted January 10, 2007 Author Report Share Posted January 10, 2007 Russ, Here is more complete information: OS X Server 10.4.7 Do you suggest that I update to 10.4.8? Retrospect 6.1.126 with Driver update 6.1.9.102. Atto UL5D SCSI Card Rom Version Released 7/21/06 (v1.52) and Atto Mac OS X 10.4.x ExpressPCI Driver v4.00. Still freezing when I go into Device configuration. That said, I am ordering a cable and terminator from Granite Digital to see if this will fix the problem. I am going to get a 2 or 3 foot cable vs the 6 footer I have now and a good terminator. John Quote Link to comment Share on other sites More sharing options...
rhwalker Posted January 10, 2007 Report Share Posted January 10, 2007 Quote: OS X Server 10.4.7 Do you suggest that I update to 10.4.8? We did, but for other reasons. It's not going to solve this problem. As with all server updates, I suggest that you clone your disk first with all services off in case the update goes badly. We had one update back when that hosed stuff, and it was the cloned disk that saved us. See my posts in some of the other forums about a RAID 1 mirror split, which is how we do it so that the server doesn't have to be taken down. Or see the white paper on softraid's site (www.softraid.com). Quote: I am going to get a 2 or 3 foot cable vs the 6 footer I have now and a good terminator. good idea. That's a bit long. We use LVD cables with ferrite beads, and terminators, from Cables to Go (www.cablestogo.com) and we've never had a problem. Good customer service, good quality, reasonable prices, fast delivery. Russ Quote Link to comment Share on other sites More sharing options...
johnkgibson Posted January 31, 2007 Author Report Share Posted January 31, 2007 :::sigh::: The saga continues.... Well... I am now going to drop 2gb of memory into the Server to see if this might help. I also didn't realize the client had ordered the system with the bare minimum of 512mb. Yesterday I was able to get BRU to backup but 3/4 of the way through I got the same freeze. After the memory I am going to try a new SCSI card and see if this is the issue. I will write more after I look into this more. Atto had me slow down the card, which I did... WEEEEE We are having some FUN HERE! Just had to get that off my chest. Quote Link to comment Share on other sites More sharing options...
rhwalker Posted January 31, 2007 Report Share Posted January 31, 2007 I'm surprised you can even boot Mac OS X Server 10.4.8 with 512 MB. You are lucky that it's not an Intel machine, because that would add another couple of Gig for Rosetta so Retrospect can be emulated. Don't see how any services are able to be turned on - certainly not mail or AFP. Simply for comparison and curiosity, our Xserve G5 (single processor) with 2 GB RAM, measured with only 1 attached user, running mail, DNS, AFP, and Software Update, shows in "top": Quote: Load Avg: 0.02, 0.06, 0.07 CPU usage: 1.8% user, 9.9% sys, 88.3% idle SharedLibs: num = 163, resident = 37.7M code, 4.43M data, 16.6M LinkEdit MemRegions: num = 9122, resident = 230M + 13.4M private, 114M shared PhysMem: 129M wired, 141M active, 493M inactive, 764M used, 1.25G free VM: 6.23G + 116M 44660(0) pageins, 0(0) pageouts Note just over 3/4 GB used, and there's currently no activity except for mail (gotta examine that SPAM..). You might want to run "top" to see what is happening while Retrospect is churning, see what the pageout number is. Note the Heisenberg uncertainty principle (observation affects the measurement - top will take up memory while doing its thing). I think you might be able to run the Chess program with 512 MB on Mac OS X Server 10.4.8, though, if the client wants to do something while waiting on the RAM. Russ Quote Link to comment Share on other sites More sharing options...
johnkgibson Posted February 8, 2007 Author Report Share Posted February 8, 2007 Russ, ATTO Tech came through with the answer... They told me to install ATTO ExpressPCI Host Bus Adapter Mac OS X Driver v3.60. This did the trick. Afterwards I was able to boost the SCSI card to full speed and still backup and restore. Retrospect is running fine now, BTW... adding the memory didn't help the freezing issue, but it did help overall performance. John Quote Link to comment Share on other sites More sharing options...
rhwalker Posted February 8, 2007 Report Share Posted February 8, 2007 Quote: ATTO Tech came through with the answer... They told me to install ATTO ExpressPCI Host Bus Adapter Mac OS X Driver v3.60. This did the trick. Afterwards I was able to boost the SCSI card to full speed and still backup and restore. ATTO support has always been excellent for us, too. I'm surprised, though, that reverting back to ATTO Driver v3.60 solved the problem. Their web site indicates that users of Mac OS X 10.3.x should be using v3.65 and users of Mac OS X 10.4.x should be using v4.00. I checked that stuff when you provided all of your version information. At the time we had our hanging issue last year with our UL4D (see above), we were using ATTO driver v3.50 and firmware v1.50. As indicated above, the hanging was corrected when we moved the Exabyte to its own channel of the UL4D. When ATTO driver v3.60 came out in May 2006 and firmware v1.52 came out in July 2006, I did not do regression testing to see if the hanging repeated with the previous configuration (Exabyte on same channel as our OS volume RAID 1 mirror) because it was too painful to recable and reconfigure (the cable management arms in our rack, while nice to enable the Xserve G5 to slide in and out, are a pain in the butt to recable and reconfigure when everything is laced down pretty) and to endure the crashing simply for curiosity, but I did test thoroughly before deployment of that code to verify that the hanging did not reappear in our configuration then (with the Exabyte on its own channel) and it has not reappeared in that same configuration now using the v4.00 driver, which was released in October 2006. It's a bit surprising that they had you revert to an older driver version when their website indicates that you should be using v4.00 with 10.4.x, which is why we upgraded to that version. Maybe a version 4.01 is in the works..... Glad that you got things going, though. Russ Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.