dimes007

Oh yeah. I wasn't worried about the speed of the E cores. I'm afraid that isolating core 0 would somehow mess up the OS or that unraid would still start some base processes on core 0 because it has to. I should have been clearer.

I'm with you: I typically isolate higher cores for vms and core 0 and its hyperthreaded pair are for unraid (at minimum). It was my intention to let unraid and dockers have all of the E cores and use all P cores for VMs. Looking at some posts here it looks like the P cores are at the start at 0. Can I isolcpu cores 0-5 [and their pairs in a 12600k] and let unraid use ONLY the E cores meaning it won't have core 0 anymore. Thanks.

Squid: Thanks for all of the amazing work on these. Can I throw in a feature request to have System Autofan log which fan it is adjusting? If you want to go further allowing us to add friendly names for each fan such as: "CPU Fan", "Top Exhaust", etc. Thanks.

Thanks for the advise. I unplugged the physical LAN NIC and went for it. So the LAN nic in pfsense is now vtnet0 (br0) passed from unRAID. as of now still using virtio but pfsense hasn't had any trouble seeing it on boots. WAN nic is still the physical x1 intel nic passed through. DHCP is working on LAN through virtio. To be clear I'm passing unraid br0 through to pfsense. I'm not passing br0.XX for tagged packets because I don't really want separate virtual nics in pfsense. my vlans are already defined in pfsense. I want all br0 traffic, even tagged packets to get to pfsense on the same virtio interface but maybe what I'm trying to do isn't possible with unraid implementation of vlans and I need to pass each vlan as a different nic to pfsense.

Hey Grid. First of all thanks for all the videos. I watched the first pfsense sense video but ventured out on my own before the 2nd was released. I'll check it out now. This weekend past I had my first taste of pfsense and VLANs (in general I'm good with unRAID, unifi and VMs). After about 3 days of effort between premise wiring, pfsenseVM configuration, netgear switch, unraid VLANs and unifi controller (in a docker no less) things are going well. My setup is as follows: PFSense has the two physical NICs passed each with 1 port. 1. WAN from cable modem. 2. Original SSID and my existing items still on 192.168.147.1/24 LAN. Other interfaces are: 3. VLAN10 is at 10.10.10.1/24. It has its own SSID as well as a guest SSID with a captive portal through the unifi controller. 4. Virtual interface is one of the virtual bridges in unRAID but as of now IS NOT USED in PFSense. Now that things work and are settled down the remaining question for anybody is one of efficiency/optimization. The physical LAN connection to PFSense has my main LAN untagged and VLAN10 tagged. The physical LAN connection to unRAID has my main LAN untagged and VLAN10 tagged. You see where this is going... I can save a switch port, gain a PCI x1 slot back and maybe gain some speed if I eliminate the physical LAN NIC and pass through the VM unraid br0 (or maybe BOTH unraid br0 and br0.10) to pfsense. I would think the virtual 10gig network is hella fast. Am I asking for trouble here? Again, this is my first experience with VLans and my first experience with pfSense so I'm not sure if I should just leave well enough alone. What do ya'll think? Thanks, --dimes

Everytime CA Backup stopping and starting dockers affects nextcloud I google "unraid hulk smash" to get back here to fix it. I may never remember how to get a shell prompt in my docker and then do # mysqld --tc-heuristic-recover commit I may never forget: "hulk smash" Thanks chaosratt.

Like Endy I'm seeing a very long run of the clear disk script. But stats are updating when I'm running so I'm certain the script didn't hang. I'm currently running the clear script on a 4 year old Hitachi 7200rpm disk on md1 and sdf. The drive isn't failing and throws no smart errors. There is no substantial activity on any other mdX devices. I turned on turbo write in the GUI (but didn't reboot if that matters). My stats:Unmounting Disk 1 ...umount: /mnt/disk1: not mounted [ed: don't worry about this. I killed it once because I was moving other data and figured I'd wait before I run it]Clearing Disk 1 ...66783805440 bytes (67 GB, 62 GiB) copied, 52920 s, 1.3 MB/s So I'm looking at about 18 days of total run time? I see read activity for all of my array drives on sdX of about 1.3 MB/s including the drive I'm clearing. I see write activity for md1 of the 1.3MB/s. I also see read activity for sdf, the drive i'm clearing, of about 10MB/s. Removing a cleared device and replacing it with a precleared drive and not having to lose parity (especially since I have only 1 parity disk) seems like a no brainer. I'm uncertain why all sd? devices are read to maintain parity. I always thought that unraid only did an XOR calc for parity on data change. I guess because its block level with DD on an unmounted drive it needs a full calculation using all disks? Is it possible to modify the script to write 0's to a file in the file system. Then unmount and zero the partition info with dd using the slower method over just the partition info? Thanks.

Being accustomed to vmware I expected ./domains/vmname to contain not only .img files but the configuration xmls as well. I have the usb flash used to boot unraid and the data disk that all of the data was contained on. Where can I find/copy/extract/export the xml files from the flash and disks of an unraid tower that isn't running. Understood that some edits will have to happen once the xml files are moved to new hardware. My biggest issue is just that I don't remember if I used OVMF or Seabios and I don't want to mess up the HAL in Windows and end up blue screening a known good and working install. Thanks. --dimes

I do note the start of the thread suggest using OVMF but I'm fully invested in SeaBIOS vms... Does CPU Pinning with SeaBios i440-2.3 work at all? For those with working CPU Pinning when running a CPU benchmark do only the pinned CPUs ramp up in Dashboard? For a dual socket config would you recommend emulator pinned on the first pair of the socket of the pinned cores or keeping emulator pinned to the first few cores of socket 0? Since I've got 16 cores w/o HT should I dedicate a core-pair (or even a core-pair per socket) to emulator pin? ======= Longer read: Life was pretty good. I was using cpupinning without isolcpu and without emulation pin but noticed that when I ran a CPU Cinebench test all cores lit up and that always irked me. Either CPU Pinning wasn't really working OR the emulator was using way more processing time than expected. Trying to sort it all out now but not having the best of luck. Windows 10 VM is down almost 40% in CPU Cinebench CPU test (I suspect this may be due to the Hyper-V sockets/cores/threads setting which I've put back to 1/4/1 but Cinebench test still shows the system thinks its 1/2/2)?. The Windows 7 and Server VMs are all within 10% which may be a good result considering emulator pin, under multiple stresses I'd hope to see overall better performance. Another note about Emulator. In that Windows 10 VM with 4 cores while running a CPU Benchmark my processor shows 25% utilization. My system at idle is at about 4%. Maxing out 4 cores should be another 12.5% so it appears the emulator is using ~7.5% of CPU time. The emulator working this hard seems to correlate with the behavorior of all cores lighting up when running a Cinebench CPU Benchmark. --dimes Lenovo D30 2 x e5-2670 v1 @ 2.6Ghz 96 GB RAM Supermicro AOC-SASLP-MV8 Intel® SSD DC S3500 Series 800 GB Cache 11.5TB Protected Storage 2 x gigE, on MB NEC uPD720200 Passed USB 3, on MB AMD Radeon HD 8570 Passed GPU Nvidia NVS 295 Passed GPU

My new unraid box is a refurb Lenovo D30 Workstation. It has: Dual Xeon E5-2670 (and 800gb ssd, 128gb of ram. Not a ton of drive bays but for ~$1350 I'm so stoked!! but I digress) There are resources shared (L2 Cache?) between the two logical cpus (hyperthreading) per core. Does it make sense to pair these "sibling" logical cpus when allocating resources to docker images or vms? On my unraid dashboard these siblings are paired under cpu speed: "core 0 / 16 1589 MHz 1199 MHz" "core 1 / 17 1465 MHz 1499 MHz" Not sure if this pairing is intentional to show they're siblings or was just done to make the display nice. To me, it makes sense to pair the sibling cpus for a given workload. So for a 4 logical CPU VM I'd take cores 14, 30, 15, 31. Thanks.