January 7, 20197 yr I have the following hardware: Case: Norco 4224 with three 120mm backplane fans controlled by the 'Fan Auto Control' plugin Motherboard: SM X8SIL-F 2x SAS cards: AOC-SASLP-MV8 8-Port (Marvell) CPU: Xeon L3406 (30W TDP) RAM: 4GB ECC Cache: 1TB Samsung EVO 860 Drives: 19 including parity The Marvell controllers have been wreaking havoc lately. I see that the Dell H310 cards are popular but they run hot. In order for all drives to be accessible I need three of them or one plus an expander. If I go down the three track I see that there is a 3D printable 40mm fan mount here: https://www.thingiverse.com/thing:2596851 But this will result in the cards blowing hot air on one another. Alternatively I can get this ventilated bracket and increase the minimum speed on the 120mm fan wall. https://www.ebay.com.au/p/Standard-Full-Heigh-Bracket-for-LSI-9260-9270-9271-9261-9240-H310-Two-Screws/1150016057?iid=172963936274&chn=ps Another alternative is to only purchase two H310s and leave the PCIe slot between them empty. Then I would need to reduce the number of drives which is not a big deal as there are four 1TB drives and two 1.5TB drives with 54% total utilisation of the array. Note: The CPU load on two of the four threads (visible on the dashboard) can flat line intermittently. When I run htop they are no where near flat lining which is confusing. When this happens the unRAID GUI remains responsive as do the Dockers but the log does not load when I click the link. This is happening right now as I am running unBalance to clear the last of my reiserfs drives. I am willing to concede that I may be trying to do too much with this server (Deluge, Jackett, Sonarr & Radarr dockers) but I want to start by replacing the problem components and see if it comes good. If I need to replace the motherboard, CPU & RAM at least I have the right SAS controllers to move forward. Can someone advise if I am headed in the right direction or is there a more elegant solution.
January 11, 20197 yr I was dealing with similar issues and it was mostly resolved by updating most containers with CPU pinning to any but the first core.
Archived
This topic is now archived and is closed to further replies.