Recent content by fermulator

More experiments (wasn't as scientific and diligent with note/result taking) TLDR: zvol configuration/tuning was not the main culprit. (Though I'm sure I could refine and tune it according to the workload of a given VM); Rather, it is the cache & IO-mode of KVM/VirtIO. `~<5-20MB/s, -->...

Last week the SSD died fwiw :) -- at first it couldn't be automatically imported on start, had to force the import since the SLOG had failed; however there was no data loss because the previous shutdown/restart was clean (and the system is running on a UPS); Back to the pool without the SLOG...

1 year later update on the Kingston SSD SoH: ID# ATTRIBUTE_NAME FLAG VALUE WORST THRESH TYPE UPDATED WHEN_FAILED RAW_VALUE 1 Raw_Read_Error_Rate 0x000b 100 100 050 Pre-fail Always - 0 2 Throughput_Performance 0x0005 100 100 050 Pre-fail...

hm! so what possible data loss are we talking? My understanding is that even in a fault scenario, non-mirror'd SLOG, zfs will detect it and kick it out without any data loss. So even using a "cheap/bad/cruddy" desktop SSD is fine for low workloads if it meets your performance needs. But I MAY...

@Rand__ & @gea; Thanks for the input on the hardware. I had the old desktop Kingston SSD just lying around, I did not (and still don't) see any problem with using it until it dies. (otherwise I had no other purpose for it). It's just for a home lab, personal VMs, "relatively" low continuous...

TLDR of verdict: 1. shoved in an older Kingston SSD as SLOG 2. enabled sync=always on all ZVOLs for VMs 3. accept risk of power loss / crash at same time as SSD failure as system is on UPS battery backup $ sudo zpool status zstorage pool: zstorage state: ONLINE scan: scrub repaired 0B in...

Of note, I still don't know what is the RIGHT configuration/setting on the ZVOL side (per all the notes above) in terms of volblocksize, and what the best physical/logical sizes are for the VM partitioning.

That is a _great_ and informative post thank you for sharing! It most certainly helps to understand a bit better about the sync-related considerations throughout the stack. Can confirm that all my VMs are not employing the blind "100% always sync" strategy, (only one of them has it enabled for...

RE: > seq write is using a QD of 4 and 16 jobs so it's scarcely any different from a random write , do you mean it is _very different_? (TBH I'm not sure about the QD and parallel job count and alignment with "typical" workloads or not) RE: > don't have a lot of IOPS to play with here and I...

Hi all experts! I would like to share my setup, and ask for guidance/advise based on what I have. I've been running this configuration for many years now, but finally looking to address the slow I/O performance in my VMs. Have spent 10-15 hours running various augmentations & tests in a...

All thank you so much for the detailed responses! Very helpful. I have opted to go for 6x of these right-angle 60cm SFF-8087 cables * Right Angled Mini SAS 36pin SFF-8087 to 8087 Data Raid Cable 60cm - Newegg.ca Given the comments the right/left angle is best, and 100cm is too long, 60cm...

Hello! Despite the mix of failures and success with users, the price for these Norco RPC chassis cannot be beat, so I've purchased one on order. Hoping to catch the eye of someone who knows the newest build for this chassis. Various other users and "the Internet" have reported strange...

Great article! https://www.servethehome.com/sas-sata-cables-guide-sff8087-8088-8470-8482-8484-single-device-connectors/ Suggest an EDIT to include SFF-8644 (noted after learning due to Norco's DS-* product line in...