This article describes a situation where the Synology 720+ NAS has poor write performance due to a disk issue. However, the NAS is not reporting an issue. SMART is reporting that everything is OK.
First, the symptoms. The following two screenshots show the actual speed when the disk was in a poor state, and the speed when the disk was replaced. Hint, the left-most image is the poor state.
 |
 |
The following screenshot shows incorrectly that everything is healthy. There’s 4 disks. 2 are for storage, and 2 are for caching. The caching disks can be ignored for this scenario. Both the storage and caching disks are in RAID1 configurations.
The following shows that disk 2 (yellow) is working at 100% utilisation, while disk 1 (red) is working just fine at a much lower percentage of utilisation (the red line mid-test was the measuring point – the cursor position, and not part of the test).
Note: The graph below was captured while large files (a few GB) were being copied to the NAS using SMB over a 1Gb/s LAN link.
The following shows the Synology doing a performance test of disk 1. It looks fine. Notice the “throughput of 165MB/s”.
The following shows that disk 2 has a serious write speed issue. Notice the “throughput of 4MB/s”. This isn’t reported as an issue, which seems strange.
The results above show that disk 1 is writing at 165MB/s, while disk 2 is writing at just 4MB/s. I am not sure why this is not reported as an issue by SMART, but regardless, if you’re seeing something similar, my solution was to replace disk 2. It’s a low-cost disk.
For completeness, the NAS details are:
- Synology 720+ NAS.
- Western Digital WD Red Plus 4TB 3.5″ NAS HDD SATA III NAS Hard Drive 5400 RPM 256MB Cache (WD40EFPX).
- Disks are in a RAID1/mirror configuration. BTRFS format.
In addition to the information relating to the intent of this article, I’ve included the following screenshot to show the rebuilding process in terms of utilisation. I find it interesting that the old/existing disk has higher utilisation than the new/replacement disk during the rebuild process. There was very little load on the server at the time of this screenshot (almost nothing at all happening). This may be typical, I just haven’t noticed it before.
