Joseph Clark

I posted about this problem in another section, but I got no replies. Thought I'd try here. I've been happily chugging along for some time with 2 UnRAID boxes, named Tower and Tower2. Recently, Tower2 vanished and I had no access to it from Windows. After a reset, things seem to boot normally (bzimage, bzroot, etc), but the system never appears as a network device as my other UnRAID Tower box does. When it boots, it appears with a "Tower: Login" prompt. I'm fairly sure this used to read "Tower2: Login." I shut down the other UnRAID box and tried to access the inaccessible box as Tower, but Windows still can't find it. There's no drive activity showing in the system after a boot, so no parity check is taking place. The flash drive appears normal, with all the files apparently intact. I'm including my syslog file here, but I don't have much experience deciphering what it means. I do note near the end that some of the shares (such as "Our Lady of Sorrows" - a church project I worked on) can't be found. Can someone advise me what to do to troubleshoot?

Update time: I received my 3 PCIe SATA cards (2 2-port PCIe x1 cards and 1 Adaptec 4-port PCIe x4). Here's how I've staggered them: Cage 1: mobo - sil3132 card 1, port 1 - Adaptec x4 port 1 - mobo - Adaptec x4 port 2 Cage 2: JMicron1 - mobo - Adaptec x4 port 3 - onboard sil3132 - mobo Cage 3: sil3132 card 1, port 2 - mobo - Adaptec x4 port 4 - JMicron2 - onboard sil3132 (eSATA) (unused at this time) Parity: sil3132 card 2 (parity is on this card by itself - the second port is unused) Cache: IDE controller (by itself - no other IDE drive in the system) The 6th (and last) mobo port is not used, and I staggered them so that a mobo port would be accessed for every third drive. The rest I staggered fairly evenly among the onboard Sil3132 ports, the JMicrons and the PCIe card ports. For parity checks, I went from just under 20MB/sec with the 2 port multipliers to just under 55MB/sec with the separate PCIe SATA cards (although this number fluctuates a bit, depending on when I hit the "Refresh" button - not sure why that is). This is near the start of the parity check. The system also boots considerably faster with the PCIe SATA cards installed than the PMs. I think it might be fair to assume that once the parity check has passed the 500GB drives in my system that I may be close to the 90MB/sec that Jim is getting with his Abit AB9 Pro mobo. That's based on the dramatic improvement I got when the parity check passed those 500GB drives when I was using the PMs. I didn't have the patience to try a parity check using the 4-port PCIe Adaptec card and one PM (without the other PCIe x1 SATA cards). I imagine it would have been a little slower (maybe considerably), but I won't know for sure. Still troubling is the mobo wailing. That hasn't changed, even though I fiddled with the CPU fan power pins. The CPU fan is working fine, BTW. If anyone has any other ideas about what this might mean, I 'd really like to hear it. So far, it doesn't seem to have had any effect on the system, but it makes me nervous to know that an unresolved issue is hanging there. Thanks, again, everyone, for all the assistance in moving my UnRAID box to another motherboard. I learned a lot in the process.

Both PMs have only 4 drives attached, not 5. One port on each is not used. Could this be what's causing the reference to 1.5Gbps? Seems like it shouldn't be a problem, right? Thanks for checking it out.

OK, here's the syslog. I found 2 references to 1.5Gbps but wasn't sure how to interpret them to find the drives.

Even if I had the time to test a PM with each of the new cards (i.e. one PM per card), I don't see how I'd stuff all this in the box. One problem with the PMs is that they have the SATA ports along the flat sides of the cards, instead of the end/edge. The SATA cables protrude into the next PCI slot space and make adding other PCI cards very tough, so there's not enough physical space for 2 2-port cards and 2 PMs in the slots on the mobo.

Thanks. The info scrolls off the screen before I can see all of it, so I'm not sure what's in the off-screen data. At least one of the hard drives, apparently, is detected at 1.5 Gbps. All the lines I can see contain "ata3 (or 4, 5, 6, etc.): SATA link up 3.0 Gbps (SStatus 123 SControl 300)," except for one that says "ata 2.05: SATA link up 1.5 Gbps (SStatus 113 SControl 320." Is there a clear way of telling if a drive is 1.5 or 3.0 Gbps, from the label on the drive? It's possible that an older 500GB drive sneaked in at 1.5Gbps, but I don't remember it. The last 8 drives have "ata3, ata4, ata5, etc., while the earlier drives have "ata2.05, ata2.04, ata2.03." I'm guessing the PM-attached drives have the ata2.05, 2.04, 2.03 designations?

When I type grep 'SATA link' /var/log/syslog I get a "Password:" prompt but am locked out on the keyboard. Nothing I type shows up on the screen. What am I missing? I got the 2 PCIe 2-port cards from Newegg, but the 4-port Adaptec isn't coming until Monday. Thought I'd replace one of the PMs with the 2 cards an do a parity check to see how much of an improvement it makes. When I get the 4-port Adaptec Monday, it should provide a better basis of comparison for using port multipliers vs PCIe cards. One thing I've noted during boot is that the process seems to take a lot longer than it does without the PMs. The system goes through a long series of soft and hard reset failures before it seems to recognize the drives as 3G.

No, I didn't capture a syslog after each of the experiments. Sorry.

OK, here's an interesting twist. I just checked the progress on the parity check and suddenly the speed has jumped to almost 60MB/sec. Estimated completion time went from nearly 20 hours at the beginning of the check (after about an hour) to roughly 11 hours at the end. I'm hoping the error sound is something simple, since the system shows no signs of instability after close to 12 hours of drive activity during the parity check. No overheating - nothing looks (or smells) out of the ordinary. I'm wondering if the extraordinarily long parity check estimate was the system predicting on the basis of all 1.5TB drives, and when the check passed the end of the large number of 500GB drives, it went a lot faster. The bottlenecks would be gone when just the parity and the single 1.5TB drives are active. This is a very good sign, I think. I plan to get the PCIe cards and hopefully the bottlenecks will be gone or dramatically reduced. I saw the $20 cards at Newegg: http://www.newegg.com/Product/Product.aspx?Item=N82E16816124004 I still haven't found a $100 PCIe x4 card. Jim? Here's the plan, when I get these cards: Cage 1: mobo - Sil3132 card 1 - x4 card 1 - mobo - x4 card 2 Cage 2: Sil3132 card 2 - mobo - x4 card 3 - x4 card 4 - mobo Cage 3: JMicron1 - onboard Sil3132 - mobo - JMicron2 - onboard Sil3132 (eSATA) Parity: x1 Sil3132 card Cache: IDE controller This gives the parity drive its own x1 lane, and it spreads out the other drives among the controllers so as to avoid congestion. My tests definitely showed that staggering the drives improves performance dramatically, and this staggers them even more effectively. Thanks, everyone.

I tried just the 6 ICH ports and the parity drive by themselves (PM drives disconnected altogether) and the results (about 20MB/sec) were pretty much the same. Not a good sign, huh? Maybe the warning sound is an indication of a bigger problem.

I'm not using any of the IDE drives for data. I transferred all that data to the new SATA 1.5TB drive. The only IDE drive in the system now is the cache drive, which is connected to the onboard IDE controller. It doesn't seem to have any affect on the system one way or the other, according to the tests I ran, one with it on and the other with it disconnected. I'm assuming it's pretty much ignored in terms of most operations, except at the point the files are being transferred at night. I think the big problem might be that I have the 2 port multipliers in the internal and external Sil ports, so all those drives are probably sharing the same bandwidth. They are staggered this way: cage 1: PM1-mobo-PM2-mobo-PM1 cage 2: mobo-PM2--mobo-PM1-mobo cage 3: PM2-mobo-PM1-JMicron1-PM2 Parity - JMicron2 Cache - IDE PM1 is on the mobo Sil connector and PM2 is on the external eSATA connector. The last slot in cage 3 is unused at this time, so there's room for one more drive. The JMicron ports are being shared by a single data drive and the parity drive. I tried to isolate parity from the others as much as possible. I've checked all the Seagate drives to make sure they're not jumpered for 1.5Gb/sec, so that shouldn't be a bottleneck. I shouldn't have gotten the PMs at all. I think I'll go Jim's path and add 2 PCIe 2-port controllers and 1 PCIe 4-port controller. The PMs I'll sell over at AVS (or here, if anyone is interested). Could you guys link me again for those controllers? I'm just not sure which Adaptecs to get. The Adaptec 4-port that you referenced, Jim, I can't find. I did find a couple of pretty inexpensive 2-port controllers at Newegg, but the 4-port units were quite expensive, not the $100 I thought we were talking about. One thing that this whole experience has made clear, though, is the ease of migrating drives from one UnRAID system to another (once I sort out these issues). I've had a lot of trouble in the past trying to move arrays created on various onboard and IDE RAID card controllers. That was one of the things I read that was so attractive about UnRAID, and that certainly has proven to be the case with this project. I haven't lost any data. Jim, I'll check the CPU fan lead when this last parity check finishes. Hopefully, it's something that simple. Again, if you guys could link me to the PCIe SATA controllers you're using, that would be a great help. I know how you feel, Weebotech, about adding those latest drives. Once these problems are solved, I'm going to be tempted to triple the size of this UnRAID box by upgrading to more of those 1.5TB drives.

I will post a syslog ASAP and check the system health in BIOS. Kinda busy, so this is lower priority for me. I meant Abit, not Asus. Most of my boards have been/are Asus - just a brain cramp. It's Abit I haven't been able to get hold of. (The only other Abit board I've owned is the one I'm replacing this one with - the Abit IC7 Max3.) I did a parity check with 6 data drives (ICH) and the parity drive (on the internal Sil3132 port). Check time was still about 20MB/sec. Interesting that the parity create time is different than the regular check. Do your remember what the parity create MB/sec time was for your system, Weebotech? And the time for just a regular parity check? Of course, my times are for parity create. For anyone reading, what are your parity check averages in MB/sec? I reset BIOS a couple of times and it did nothing for the wailing. I haven't updated the BIOS. I'll try that. I'm using the internal Sil3132 port. I got the PMs, though I'm sorry I didn't just get the extra controllers. On my old Abit mobo, I think the parity checks were significantly faster, for the same amount of data - about 40% faster, IIRC. I was using 2 PCI SATA cards (no PCIe slots on that board) - a Promise 4-port and an Adaptec 4-port. The onboard SATA controllers were just 1.5Gb/s, too. That system was all 500GB drives, the same ones I'm using now (except, of course, for the 4 IDEs, 2 on each of the onboard IDE controllers).

An update: I still don't have a solution to the problem of the mobo warning wail. I've tried numerous times to call tech support at Asus, but no one ever answers the phone. It's always been busy. I tried to log on to their support forums, but I never get the final piece of the puzzle - an email that allows me to verify that I'm registered. They say they'll send me the email, but it never comes. I've tried parity checks with the port multipliers and without. The IDE cache drive that I'm using doesn't seem to affect the parity check speeds at all, so I think I'll be OK using it instead of buying a SATA drive for cache. The bad news is that parity check speeds are awful, with or without the PMs. Even if I use just the mobo SATA controllers, without the PMs, I'm only getting about 20MB/sec. The drives are a real hodge podge (Seagate, Hitachi, Maxtor, WD), but they're all 500GB (or 1.5TB), 7200rpm SATA drives. With 15 drives in the system, it estimates about 20 hours for a parity check. Not good. Staggering the drives does make a remarkable difference in the estimated parity check completion times, reducing the time by about 1/3. The first time I did a parity check estimate, it guaged it at over 28 hours. That's with 13 500GB data drives, 1 1.5TB data drive and a 1.5TB parity drive. I've let the parity check run for at least an hour, since in the past it's been pretty close to accurate for a check after that amount of time. I have to assume, since I get this error sound from the mobo, that whatever is causing the problem is a likely suspect to be causing the awful performance as well. I'm very frustrated, though, that I can't get Asus to find out what this error sound indicates.

I'm using a stock HSF with an Intel E6400 CPU. I was concerned about using both Sil3132 ports, if they share the same bandwidth. If the staggering of drives concept works, though, it could still work out OK. With the 2 PMs on the Sil ports, I could stagger them like this: PM1-mobo-PM2-mobo-PM1-mobo-PM2-mobo-PM1-mobo-PM2-mobo-PM1-JMB1-MP2-JMB2 = 16 drives. I'm still thinnking of using one of the IDEs as a cache drive. Unless it's transferring, it probably doesn't affect the array. I can put parity on one of the JMicron ports. About the AMS Venus, Weebotech - I have two of these cages in this system. How did you test it in PM mode with the JMicron ports? Edit: I think you're talking about a different cage than the ones I have installed - ones with built-in RAID capabilities, right?

I can't find the CPU temperature threshold settings in the BIOS. I must be overlooking something simple. Help. The Addonics PM card's instruction manual says it works with the Sil3132 contoller - no mention of the JMicron. Do you guys know if the Sil3132 ports (SATA and eSATA) share the same bandwidth, or are they independent? Thanks for the feedback.