rdenney

Is that the only fault, or does pre-clear.sh do something else, too? Rick "who can live with this if it's a known issue" Denney

Hope this is easy: After installing a new parity drive last night, I moved the old parity drive to an array position. In the Simple Features version of the WebGUI, I assigned the next position to the former parity disk. The system terminal message I got showed that preclear had begun. It had counted up to about 2% when I had to shut down my web browser for another reason. Now, when I try to use "tower/main" as a URL, my browser can't find it. I can, however load unMENU, which reports that the array is stopped and that parity is not valid: disk new. This what I would expect. The last line in the syslog indicates that preclear has been started. And when I refresh the unMENU Main page the writes for that disk shows to be counting up. That tells me that the preclear is proceeding. But I can't seem to restart the main webGUI. Even when I click on "unRAID Main" in unMENU, the browser times out and reports a time-out error. Is this normal? Should it be? My gut feeling is that emhttp invoked the preclear, and while that preclear process is underway, it goes deaf. Rick "really just curious" Denney

Yes, though there is nothing on the RMA form or web page that indicates such. Once I ordered the label, it appeared I would have to wait for it to be shipped to me, which seemed insane, so I called them and waited on hold for about half an hour. They cancelled that shipping label and they are sending me a new one via email, at their cost. So, at least it's a free pain the butt rather than one I have to pay for. The (so far) good drive is building parity, with 408 GB done at a rate of 95 MB/s. I'm not tempted to complain about that! Once this drive has built parity, I'll install the old 2T parity drive as an array drive. Rick "who'll have to wait a couple of weeks to mess with the replacement drive" Denney

The two that I bought were packed in plain white boxes, both put inside a larger box and "prevented" from moving with crumpled paper. The inner boxes were fine, but one of the drives was DOA. It offered up a mechanical dragging noise, various unhealthy beeps, a resigned sigh, and a refusal to communicate with my disk controller. The other drive, when put in the same position, is now building parity at about 100MB/S. One thing I don't like about Newegg, and that's having to return something. Pain in the butt. I also don't like having to pay to return something that is broken right out of the box. Grrr! Rick "grumble, grumble" Denney

I bought an M1015 flashed already from an ebay seller (daiv-2012), and he also included the necessary cables for 8 drives. Make sure you have it flashed with the LSI9211-8i (IT) or LSI9210-8i (IT) firmware. By purchasing it with the firmware already flashed as is possible with this seller, you avoid most difficulties with motherboards. Some motherboards that have only one PCI-e 16-lane slot intend it to only be used for a graphics card, and will not accept anything but a graphics card. That was true in my case, with a Dell Dimension 4700 that is now half a dozen years old. But when I flashed a new BIOS into the motherboard (using Dell's final and latest download), that opened up the PCI-e slot for multi-purpose use. If your motherboard was early in the PCI-e deployment cycle, you might make sure you can update the BIOS if necessary. I'm expecting that newer motherboards won't have this issue. But there is no way to know whether it will work until you try it. Sorry, but that's as good as it gets. That sort-of comes with the territory with a home-built computer where you are re-using old stuff, in my experience. These are good controllers because they provide an independent PCI-e lane for each disk. That removes the PCI bus from any chance of being the bottleneck in the system. They also seem to be built for the enterprise market. The cables and connectors are better than usual. You will need a late release candidate of unRAID version 5 to use any of the LSI controllers. I'm using the latest, rc8a. I don't see any M1015's for sale on ebay at the moment, however. Rick "noting that the M1015 in this box is faster than the on-board SATA ports" Denney

Actually, it is easy to determine "stable for whom". When all the requirements for a particular release have been fulfilled, and remain fulfilled, then it is stable. When it is unstable while attempting to fulfill an unstated requirement, that is a matter for a future release. The real argument comes in determining what requirements must be fulfilled. And here we have what seems to me a key part of the issue. Mr. Lime-Tech apparently wants features that would make his upcoming hardware product more effective, and those use cases are part of HIS concept of operation and he therefore has requirements related to those uses. But nobody here is using that hardware product, so those features may not be driven by the use cases of current users. Of course, the requirements here are not really documented very well or at all, so it does open the door to debate about what the current use cases actually are. That's a more crucial argument that the one about whether a release candidate can get a new feature or not. I certainly want a new feature in a stable release--I want the r8169 driver back in there. My suspicion is that the hardware product has consumed more resources than was expected, and Mr. Lime-Tech is possibly hoping for a change of fortunes to make the progress he desires with that product. Perhaps he thinks that will happen momentarily, allowing him to focus back on issues affect some users here. Maybe that outcome has been elusive, and if related to a new product, he might not want to bare his soul on this public forum. (He is still alive, at least. I got my license in a day.) There is a free version of unRAID so it is indeed possible to fully test the product's main features before committing to making a purchase. That's what I did. When I was sure I'd found a way to support my hardware, I went ahead and bought the license. For me, weaknesses include leaving off the r8169 driver from the latest release (the one that replaced it does not work with cheapie 8169-based NICs), and documentation that assumes everything goes swimmingly and provides little explanation of why things are so one can have a direction if they don't. The second is not uncommon, even in products that cost orders of magnitude more than unRAID. The first was fixed by a forum member, though it requires a kernel not downloadable from Lime-Tech. I had to use the latest release to provide hardware support for my hard disk controller, which is pretty high performance and now pretty cheap but which was not supported until recently. As I said before, I'm here because NAS4FREE had issues I could not resolve, and their support forum was too inactive to get reasonable support. Support here was terse but it got me pointed in the right direction so I could learn enough to solve my issues, and I can live with that. NAS4FREE attempts far more than unRAID, with even less general support. But, hey, it's free! But my use cases are easy-peasy. I only want a file server, not a media server, so all the media serving applications that concern many here will never get installed by me. That's the danger of assuming one's own success should be transferable. What would help is a clear set of use cases and requirements, at least at a high level, to guide version updates and configuration management. I gather that's what the Roadmap was for, but it takes effort to keep that up. It's easy to start a sentence with "somebody should..." Rick "who does systems engineering for a living" Denney

Thank you--that description (and the thread you linked) was what I was hoping to learn before I started guessing and pushing buttons. I was able to test all the elements and be sure I had addressed all your "maybes". I have no data on the drives as yet, so there was no harm in conducting a bit of dangerous testing to make sure. I pressed the power switch. Sure enough, it appeared to go through the clean powerdown, and a check of the log that it saved in the process confirms it. I then tested the UPS. APSUPSD only reports battery charge, time left, and load percentage realistically when the status is ON BATTERY, but the numbers I saw then were completely plausible. And after 300 seconds, it did an orderly shutdown. This time, the log announced the power failure, and 300 seconds later the log started to look identical to the log created when I pushed the power button. Rick "very close to putting this in production" Denney

Okay, more exploration. In another thread, Joe mentioned that the following should be done when installing the clean power down in the go script: CTRLALTDEL=yes installpkg /boot/packages/powerdown-1.02-noarch-unRAID.tgz [ -f /usr/local/sbin/powerdown ] && mv /usr/local/sbin/powerdown /usr/local/sbin/unraid_powerdown [ -f /usr/local/sbin/unraid_powerdown ] && sed -i "sX/usr/local/sbin/powerdownX/sbin/powerdownX" /etc/acpi/acpi_handler.sh [ ! -f /usr/local/sbin/unraid_powerdown ] && sed -i "sX/sbin/init 0X/sbin/powerdownX" /etc/acpi/acpi_handler.sh sysctl -w kernel.poweroff_cmd=/sbin/powerdown He explained that the next-to-last of these changed the ACPI_handler script to replace an "init" command with a "powerdown" command, and the previous lines were used to replace the unRAID-supplied powerdown (which is renamed "unraid_powerdown") with a clean powerdown. After a fresh reboot, I have looked in /etc/acpi/acpi_handler.sh and found where the change he described was made. And I see "unraid_powerdown" in /usr/local/sbin. UnMENU reports the clean powerdown is installed, and before rebooting I set it to automatically reinstall on boot. I don't see where the final sysctl command is issued, but I suspect that is in an install package somewhere, perhaps for APCUPSD. I think the next step is to press the power button and see what happens. Rick "film at 11" Denney

Joe, thank you for moving it. I was unaware of the protocol. Anyone have any thoughts? Rick "anyone?" Denney

My principle use case is providing an online destination for a ShadowProtect backup process running on each of two machines in my home network. Incremental backups to the NAS are performed by ShadowProtect daily, with complete image rebuilds monthly. Both workstations are used for photo editing. Working files are always stored on workstation disks, so backup is only accessed for reading when some recovery is needed. One machine's backup image file is about 400GB, and the other's is about 300GB. Secondary use case is to provide a guest-accessible storage spot for general file sharing of a temporary nature. UnRAID server will replace D-Link DNS-323, which is a two-drive NAS with RAID1 mirroring. Maximum write speed to DNS-323 is 11MB/s, despite upgrade to 1000BaseT network connection. Speeds six times that are routine using unRAID server. DNS-323 drives were 2TB each, providing 2TB storage. It took 15 minutes just to erase old images, which had to be done routinely because the disks filled up so often. Internet access provided by broadband cellular, which is limited to 8GB/month, preventing any cloud backup option. Router is Cradlepoint MBR1000, which provides firewall for internal network (though IP address for access from outside is not stable at all). Router used for DHCP, but unRAID server given a static IP outside the DHCP range (as is also the case with three printers). Network switch is D-Link DGS 1016D. This project began with the gift of a disused Dell desktop, which I repurposed to become a NAS. Prices shown are amounts paid for the project. OS at time of building: unRAID 5.0-rc8a Plus CPU: 2.8 GHz Intel Pentium 4 model 520 Motherboard: scavenged from a Dell Dimension 4700 (free), BIOS Version 10 (the final version provided by Dell for this MB). The only useful tidbit provided with this MB was a PCI-e 16-lane slot. Everything else is pretty unimpressive. The on-board NIC was Fast Ethernet only. RAM: Four 1GB Crucial DDR2 unbuffered non-ECC sticks (free) Case: NZXT Source 210 black mid-tower case ($40) Drive Cage(s): Case contains fixed cage for 8 3.5" drives Power Supply: Inland ATX-ILG-600R2 600-watt, single 12V rail. ($45 on sale) NIC: Tenda TEL9901G (Realtek 8169 chip, requires r8169 driver). MB NIC disabled in BIOS. ($12) SATA Expansion Card(s): IBM M1015, flashed with LSI9210-8i (IT) firmware, PCI-e 8-lane, with forward SATA expansion cables ($125, with BIOS flashed already). Cables: As supplied with above, except for several sloppy-looking fan splitters ($3) Fans: Stock fan exhausting from top of case. CPU cooling system integral fan in exhaust orientation, which didn't fit the case that well but nothing a couple of self-drilling, self-tapping screws could not fix (fan and CPU heatsink as from Dell). Package of four 120mm CoolerMaster quiet cheapies, two use in front of drive stack, in intake orientation. Fan on power supply in exhaust orientation. Air is brought in from front and exhaust to top and rear. ($10) Parity Drive: 2 TB Western Digital Caviar Green ($110) Data Drives: 2 TB Western Digital Caviar Green (1), 2 TB Seagate Barracuda LP (2), total storage 6 TB ($330) Windows drive: Left in system to provide boot capability when unRAID thumb drive removed, whatever 40G drive came from Dell. Installed in 5.25" slot using adapter, and connected to slow MoBo SATA port. All 8 3.5" slots in case reserved for unRAID future use. Cache Drive: None OS Drive: Staples Relay Micro USB Flash Drive (16G), because it was microscopic and doesn't stick out the back of the server box. ($20) Total Drive Capacity: 8 drives, including parity drive Total cost of this project: ~$700, including array drives. Primary Use: Photo storage, primarily as a backup. All files are stored on disks in each workstation, and backed up daily to unRAID file server. External storage on BluRay disks made occasionally. Internet access to work area limited to Verizon LTE and thus 8GB per month, so routine cloud backup services not an option. Likes: Very quiet, runs cool, easily expanded, easily managed, far faster than previous D-Link DNS-323, cheaper than four-drive alternatives, active support community (compared to NAS4FREE, which was the alternative) Dislikes: Big honkin' case Add Ons Used: unMENU, Simple Features, APCUPSD (with installed 650VA APS Back-UPS) Future Plans: Add drives as needed Performance: Parity check 85MB/s at the end of 2TB parity check. File transfer from Windows workstation: 60MB/s typical Power consumption wasn't really the objective and thus is not measured. Cheapie box is not ugly, but it's also not compact. But drives can be removed and added to fixed cage without removing anything else. Installed drives with space between them (why not?), and fans oriented to blow through the cage from the front. Fan wires are ugly, though, and the power supply supplied wires were a little voluminous. Rear shows NIC displaying gigabit connection. Only other light is the tiny LED on the USB flash stick. Drive cage has room for 8 drives. Top drive is Windows OS drive which will boot if USB OS stick pulled. It's wired to MoBo SATA port. Array drives are wired to M1015 NIC and IBM M1015 disk controller installed. Boxes for those looking up part numbers. Rick "under $300 without drives, and quite high performance" Denney

Just use the version shown in the post that I linked. It includes the r8169 driver, and your 8169-based NIC should then work. That version is otherwise identical to rc8a. Rick "agreeing that Realtek 8169 support should be included in any release candidate" Denney

(I'm using 5.0-rc8a.) Joe L. included a detailed description of this issue in this post: http://lime-technology.com/forum/index.php?topic=2068.msg82590#msg82590 When I look at unMENU package installer, I see that the "Clean Powerdown" package is shown as "Currently installed. Will NOT be reinstalled upon reboot." I installed the Clean Powerdown feature apparently while installing the APCUPSD package, which I installed using a newer procedure that I found in a recent thread that I cannot now recall. (I have installed APCUPSD version 3.14.10, rather than 3.14.8 shown in the unMENU package installer.) UnMENU reports that APCUPSD is "Installed. Not downloaded." The APCUPSD service is shown in the Simple Features version of the regular web GUI, and it shows what portion of the relevant information that is supplied by my cheapie APC Back-UPS. These messages persist after reboots, so I'm given to believe that the Clean Powerdown feature remains installed. My suspicion is that the Clean Powerdown feature is not installed in a way that unMENU recognizes when its package installer checks for it. Is that the case? Is there a way to know for sure? And, given the answer to that, are Joe's instructions for linking the power button service to the Powerdown feature in that old post from 2010 still valid, given that some newer version of Clean Powerdown might be installed? Given the inclusion of many features into the later release candidates, it's a little hard to know which unMENU packages are now built in or otherwise superseded by events. Rick "who would like the power button to do its job, but can already do a clean powerdown using Cntrl-Alt-Del" Denney

Okay, so I'm freaking blind. Right there on the Main page was the check box to allow me to press the format button. Sheesh. I must have truly been non compos mentis last night. The format took only a few minutes and now I'm doing a full parity check. Moving about 110MB/s, which seems fast enough. Thanks for putting up with me. Rick "in business for now" Denney

Read the last few pages. I will just say this: I posted a question about a driver for my disk controller on the NAS4FREE forum two weeks ago, when I was experimenting with that system. Nice guys, but it took seven days to get a response, and then that response was to ask me how it turned out. This system provides better support. I could always have bought a Synology NAS and not had to worry about crowd support (with its ups and downs). I just didn't have the extra grand-plus laying around. Rick "who already has a DLINK DNS-323, which is polished enough but provides no knobs to turn" Denney

This is a note just as bread crumbs for others who may run into this issue. In reading the release notes, the r8169 driver has been included, then removed, then included, and ultimately removed. It has been "replaced" by an r8168 driver. I read where the Realtek driver was a consistent problem for one reason or another. When I installed 4.7, it worked with my PCI cheapie GIG-E network card that uses an 8169 chip, without issue. But I needed a late release candidate to support my LSI9210-flashed IBM M1015 disk controller, but I found that rc8a no longer includes the r8169 driver. The r8168 driver DOES NOT find the 8169-based no-name network card. One expert user (see here: http://lime-technology.com/forum/index.php?topic=22615.msg201529#msg201529) rolled an r8169 driver into the rc8a kernel and has offered that for download. I am using that build and it works perfectly with my network card. Okay, there are reasons the r8169 driver was removed, I gather. But if one of those reasons is that the 8168 driver provides the needed functionality, then it my case it did not. Every single PCI-based NIC at Micro Center, where I was shopping for bits for this project, used the Realtek 8169 chip. Nary an Intel chip in sight. Given that many (like me) are repurposing old motherboards, a gigabit NIC might not be integrated already (mine was Fast Ethernet). The NIC I used was all of about eight dollars--an easy way to add a gigabit network interface to an older board. That board is otherwise abundantly applicable to use wiht unRAID, it seems to me, so perhaps before rc8a becomes a stable version, consideration should be given to reinserting that r8169 driver officially. Rick "bringing what has been buried in other threads to a thread title for ease of searching" Denney