Hello all, total zfs/truenas noob here, we are trying to rip and replace VMWare in a quick cost effective manner and the CTO wants ProxMox and TrueNAS. Our current dedicated SAN’s are just turning 10 years old and no money to replace them now.
Plan is to start with community editions of Proxmox and TrueNAS to test the environment with lower priority vm workloads in a 3 node cluster (~ 90 VM’s). There should be a fair bit of small file I/O, I have been given HPE DL360p gen 9’s with a 24 bay disk shelf populated with all 1TB 10k sas2 spinning disks - . I am having trouble finding addin cards for SLOG and L2ARC. I can populate the truenas server with 256 - 768GB of ram for primary read cache. I am primarily looking for the best power protected slog device to use. S,o what is everyone using for SLOG and L2ARC on older gen 9 HP servers? Any hardware recommendations and experiences would be appreciated.
So first question is, have you read over the specific use cases of when a SLOG is required and should be used?
Optanes tend to be the main go to for SLOG’s from my reading.
More ram in the TrueNAS would be better than going back to L2ARC.
My next question will be, if they want to move production workloads on to this, are you just planning to use a single TrueNAS system? = Single point of failure…
have you considered reaching out to XiSystems directly to get some quotes and support? (or that will come after proof of concept?)
My first question would be: Do you mean
- TrueNAS to provide storage to Proxmox VMs? or
- TrueNAS virtualised on Proxmox?
1 is a clear case for a SLOG, and Optane or a Radian CMS card is the answer.
If 2, why?
Wish Optane’s were available new. Assuming new is the thing, that pretty much means Radian right?
They’re available new right?
To provide more detail, we are an ISP with a large LTE deployment, currently running workloads on VMWare clusters with dedicated SANs for ourselves and MSP customers workloads running on our clusters.
The end goal is to eliminate VMWare completely during a period where revenue is way down and no budget is available, if we don’t, our September ‘renewal’ with VMWare will break us.
I have our old LTE EPC core servers from last years upgrade to repurpose for compute, these will run ProxMox in a cluster , these are Dell R740 nodes (Xeon Golds with 512GB RAM).
For shared storage I have 3 DL360p gen9 servers (E5-v4) with 256GB ram available to use for TrueNAS/virt. POC testing. Current plan is to use one 360p server, a D2600 disk shelf with 48TB made up of 4TB drives, and a pcie optane card for slog with all the RAM I can stuff into it for read cache. Hoping I can get by without L2ARC.
We will not be deploying anything mission critical to it, our environment has multiple geo diverse redundant services like DNS, NTP, syslog, radius, web servers etc. that we would migrate some of the redundant VM’s to for testing. After this stage goes well, also migrate bigger platforms like Atoll RF planning software, Ciena MCP, various GIS platforms, etc. slowly added to get a gage on the performance, collective platform and build knowledge among staff.
So short version, 3 R740 servers running Proxmox clustered, using a single DL360 for shared storage on dedicated 100/10G storage network with a workload made up of mostly redundant services to start. Load testing we will use some of our RF planning and GIS systems to stress it and find limits.
Last point, I only have 2k to purchase storage and required components for the proof of concept so I am hoping to save some time and mistakes but just starting with ‘ask the experts’.
Longer term we are planning to budget for our big 5 year refresh (2026) most likely with M30 or M50 HA storage units and clustered Proxmox unless there is some show stopper which I don’t expect to see in evaluating this setup.
As @MBILC mentioned, seeing as this is enterprise it wouldn’t hurt to reach out to iX to discuss a proof-of-concept so you can also get a quote for support contracts and the like.