August 24th, 2024

The formatting of Oracle-branded HGST SSDs (2021)

The article details challenges with Oracle HGST SSDs on Linux, leading to a migration from Oracle's ZFS to OpenZFS, which resolved compatibility issues and improved performance, encryption, and compression.

Read original articleLink Icon
The formatting of Oracle-branded HGST SSDs (2021)

The article discusses the challenges faced when using Oracle-branded HGST SSDs with Linux, particularly in the context of ZFS file systems. The author recounts their experience of migrating from Oracle's ZFS to OpenZFS due to issues with file system bugs and a lack of support for non-paying customers. They encountered significant errors while formatting the SSDs on Ubuntu, which were attributed to a "type 1 protection" feature that added extra bytes of error correction information. This feature caused compatibility issues with certain tools, leading to read and write failures. The author found a solution by disabling this protection using specific commands, allowing for successful formatting and integration into a ZFS pool. They also noted performance improvements with their new setup, highlighting the advantages of OpenZFS over Oracle's offerings, particularly in terms of encryption, compression, and overall functionality. The article serves as a technical exploration of the intricacies involved in managing enterprise storage solutions and the benefits of open-source alternatives.

- The author migrated from Oracle's ZFS to OpenZFS due to support issues and file system bugs.

- Compatibility problems arose from the SSDs' "type 1 protection" feature, causing read/write errors on Linux.

- A solution was found by disabling the protection, enabling successful formatting and integration into ZFS.

- OpenZFS offers advantages over Oracle's ZFS, including better encryption and compression options.

- The new setup demonstrated significant performance improvements in data management tasks.

Link Icon 4 comments
By @somat - 8 months
It reminds me of when in the early 2000's I had bought an sgi off ebay and was scouring the local computer repair shops for cheap 80-pin scsi drives. which were understandably a bit rare in my rural area. I I took a chance on one cheap drive, I think he said it came out of some sort of ibm array, but it ended up having 510 byte sectors, Which is persona non grata in any reasonable system. And after a lot of reading the docs(sgi had some great docs) I determined the only thing to try was a low-level format. By this point ide drives had lost the ability to low level format but scsi drives were still able to do it. And after a few tense hours, the docs warn that if it failed or the parameters were wrong the drive would probably be bricked. it worked! and served quite a few happy years in my O2.

One my favorite things I found in the SGI docs was instructions on how, for the highest speed xfs array you should set it up, one drive per scsi controller then stripe it. An arrangement I found exotically extravagant.

By @cesarb - 8 months
I knew you could reformat NVMe drives to have 4096 byte instead of 512 byte logical sectors, but didn't know there was a standard command to also do that for non-NVMe drives. I might try that next time I get a new non-NVMe SSD.
By @auspiv - 8 months
I did a ZFS SLOG benchmark last year and the author's conclusion what drive it is (HGST SS300) was one of the test subjects - https://austinsnerdythings.com/2023/01/31/zfs-slog-performan...

That said, the model numbers don't really match up - which is explored a bit in the original post. Don't think a 200GB drive exists in this lineup.

Love HGST drives. Also have a SSD400S 200GB that has 18 PB of endurance.

By @ComodoHacker - 8 months
Somehow this page makes my latest Firefox continuously eat all available memory.