nextadmin Posted August 9, 2023 Share Posted August 9, 2023 Hi, My lab is using UNRAID_VERSION: 6.12.3, is based on a Ryzen 5 3400G with 16 GiB DDR4 on a asus Motherboard DS3H-CF based on a B450M chipset. ( joined the diags files ) I have unraid managing ( so far ) an array of one disk parity + 5 hdd, one ssd as a cache ( 450 GO so far, planning more if needed ), one hdd as a device for backup. Classique shares, with cache to array for most of my setup. one docker ( stopped ) for managing VM backup manualy. one Debian 12 VM running nextcloud so far ! Planning to have other VM for dev purpose. It took my several try ( dockers, vm, apps tested ) to finally have what i want. MY PROBLEM 1: - with 16GO DDR4 ram, if parity check is running, i have 85% of the ram consumes. MY PROBLEM 2: - If i upload some files to my debian/nextcloud machines, the system hang/crashe after a few minutes ( my upload was 107 GO and it takes like 10 minutes to build up the ram to more than 100% and crashes... MY QUESTION: - do i need more ram ? - how much ram is needed per VM ? - Do i need to plan for a larger SSD for cache use ? - do you have any comments on my setup MY GOAL: Apps and Dockers are NOT what i want to use with unraid. Only a few if absolutly needed. My main goal is to have a production debian/nextcloud VM running as production. And Somes others VM on a start/stop scheme when needed... still have 25 days to decide to buy a licence or find another solution. thanks. nextlab-diagnostics-20230809-2240.zip Quote Link to comment
JorgeB Posted August 10, 2023 Share Posted August 10, 2023 You are assigning 12GB to the VM, note that KVM always adds some overheard, so in practice you are leaving less then 4GB for the OS, try to leave around 8GB for the OS. Quote Link to comment
nextadmin Posted August 10, 2023 Author Share Posted August 10, 2023 Hello JorgeB I did not realize my mistake. Went back to orignal 8/8, but crash did continue... Anyway, those crash were not good. I did a mem86 and find out one of my ram was bad... Just ordered new ram to definitely fix the problem ( hope )... Thank you for pointing this out ! 1 Quote Link to comment
nextadmin Posted August 13, 2023 Author Share Posted August 13, 2023 Issue is fix. I bough 2 16GO kits... Now, memory size and quality is no more an issue... Also, i went to a double 1TO ssd in order to have redundancy for the cache array... Quote Link to comment
nextadmin Posted August 14, 2023 Author Share Posted August 14, 2023 (edited) OK, here that i am doing, thank to JorgeB who was nicely helping: Using some "old" components ( ryzen 3400G, B450M Motherboard, Exos X18 16To )... 32 GO DDR4RAM ( new, tested ok ) 32 GO USB 3.1 key ( just ordered some as i have no more spare ) HBA BEYIMEI 10 X sata3.0 6GB/s based on ASM1166 + JMicron JMB5xx ( tested ok but slow and bottlenecking the system ) 1 parity disk Exos X18 16TO ( have one sitting on a shelve for more than 4 month, time to waste it ) ARRAY of 5 HDD x 4TO ZFS ( ParityST4000NE0025, bought between 2016-2020 ). Will use them until they die, hopefully not all at once. CACHE of 2 X 1TO SSD ( redundancy ) ( just ordered ) Might be a good start. Planning to updated the HBA to a AS9300-16I 12GB/s if the pci 4.0 bios gets available... or i will also change the motherboard... ( getting expensive ) When the HDD from the data array will die, i might simply change them for larger HDD, or for fewer SSD ( 2 TO price are nice , and i will have around 8 free sata slots if i update the HBA ) I will run a Debian12/NEXTCLOUD VM from the cache, store the user's data to the ARRAY ... and backinp up the data in two ways : ZFS Snapshots on the server and a schedule rsynch to an intel nuc and HDD 10G box Edited August 14, 2023 by nextadmin Quote Link to comment
ChatNoir Posted August 14, 2023 Share Posted August 14, 2023 5 hours ago, nextadmin said: HBA BEYIMEI 10 X sata3.0 6GB/s based on ASM1166 + JMicron JMB5xx ( tested ok but slow and bottlenecking the system ) It's probably using a port multiplier 1 Quote Link to comment
JonathanM Posted August 14, 2023 Share Posted August 14, 2023 13 hours ago, nextadmin said: Will use them until they die, hopefully not all at once. I understand the sentiment, but be aware that Unraid's parity array can only tolerate as many failures as there are parity disks, be that 1 or 2. Any failures on data drives beyond that will result in data loss on all dead drives. Unfortunately drives don't always give a lot of warning before they die, especially when they are rarely used, as can be the case with Unraid's independent disk strategy where often times a disk can stay spun down for weeks at a time. It really sucks to replace a dead drive and have one of the drives you didn't think had any issues die during the rebuild attempt. If you are planning to push the drives through failure... 13 hours ago, nextadmin said: backinp up the data in two ways : ZFS Snapshots on the server and a schedule rsynch to an intel nuc and HDD 10G box please make sure all important data is constantly backed up away from the server, so drive failures don't take you by surprise at the worst possible moment. 1 Quote Link to comment
nextadmin Posted August 16, 2023 Author Share Posted August 16, 2023 Done: Definitely upgrading controller. Going for a more simple HBA 9340-8i as i realized: - I will never upgrade to SAS ... - I will surely not upgrade my array to mvne disks ... - SSD are not completely out of scope if ever i have dying disk ... - but i want to take advantage of a real PCI3.0 LSI HBA ... Find the card with theartofserver Quote Link to comment
JonathanM Posted August 16, 2023 Share Posted August 16, 2023 35 minutes ago, nextadmin said: Going for a more simple HBA 9340-8i Before you buy, please ensure that it can do true IT mode. I'm not up on all the latest part numbers, but I'd be more inclined to think a 9300 would be only IT, where a 9340 may come in IR mode, and may or may not be changeable to IT. Quote Link to comment
nextadmin Posted August 16, 2023 Author Share Posted August 16, 2023 40 minutes ago, JonathanM said: Before you buy, please ensure that it can do true IT mode. I'm not up on all the latest part numbers, but I'd be more inclined to think a 9300 would be only IT, where a 9340 may come in IR mode, and may or may not be changeable to IT. Thank you. I did check with the US vendor, and i did choose this one excatly: LSI SAS9340-8i ServeRAID M1215 12Gbps SAS HBA P16 IT mode for ZFS FreeNAS unRAID Actually, i bought a BAD USB KEY from amazon, who failed me after less than one hours using unraid ( warning message the key is corrupted or going offline )... I start to beleive i am curse ! The bad usb 3.0 is a 32GO GIGASTONE ... Hopefully, the config folder is safe... and i was using the beta unraid to save it ... I just hope i wont go to bad usb keys too much... Quote Link to comment
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.
Note: Your post will require moderator approval before it will be visible.