Harddrive Write Cache Disabled.

[tjkaz]

Can you paste the disk section from /usr/syno/etc.defaults/scemd.xml. It appears that the 4.0 firmware has different content for this file for the different models of DS...which if true will be very strange indeed.

Not sure how to access "/usr/syno/etc.defaults/scemd.xml". I presume that's a hidden file/folder on the DS.
FWIW, I do not have a RAID setup; two disks, two volumes. I'm wondering if this setting is being disabled for selected drives AND those being used in a RAID setup.

[Larry]

Can you paste the disk section from /usr/syno/etc.defaults/scemd.xml. It appears that the 4.0 firmware has different content for this file for the different models of DS...which if true will be very strange indeed.

From my ds410 with 4.0-2198.

Code: Select all

 
       <disk model="ST9120821A">APM_255</disk>
       <disk model="HTS722020K9SA00">APM_255</disk>
       <disk model="ST980811AS">APM_255</disk>
       <disk model="MHV2060BH">APM_255</disk>
       <disk model="MK6034GSX">APM_254</disk>
       <disk model="ST9100824AS">APM_255</disk>
       <disk model="HM250JI">APM_255</disk>
       <disk model="MK1637GSX">APM_254</disk>
       <disk model="ST9100824AS">APM_255</disk>
       <disk model="ST9120822AS">APM_254</disk>
       <disk model="ST9160821AS">APM_254</disk>
       <disk model="WD2500BEVS-75UST0">APM_254</disk>
       <disk model="HM250JI">APM_254</disk>
       <disk model="WD2500BEVS">APM_254</disk>
       <disk model="WD1600BEVT">APM_255</disk>
       <disk model="HTS543232L9A300">APM_255</disk>
       <disk model="WD1600BEVE">APM_255</disk>
       <disk model="MK2546GSX">APM_254</disk>
       <disk model="MHV2080BHPL">APM_255</disk>
       <disk model="WD1200BEVE">APM_255</disk>
       <disk model="HTS543225L9A300">APM_255</disk>
       <disk model="WD3200BEVT">APM_254</disk>
       <disk model="MK8025GAS">APM_254</disk>
       <disk model="ST9200420AS">APM_254</disk>
       <disk model="WD3200BEVT">APM_254</disk>
       <disk model="HTS543216L9SA00">APM_254</disk>
       <disk model="MK4058GSX">APM_254</disk>
       <disk model="WD6400BEVT">APM_254</disk>
       <disk model="ST32000542AS">APM_255</disk>
       <disk model="ST95005620AS">APM_255</disk>
       <disk model="WD7500BPVT">APM_254</disk>
       <disk model="WD30EZRX">WCACHE_OFF</disk>
       <disk model="WD25EZRX">WCACHE_OFF</disk>
       <disk model="WD20EARX">WCACHE_OFF</disk>
       <disk model="WD15EARX">WCACHE_OFF</disk>
       <disk model="WD10EADX">WCACHE_OFF</disk>
       <disk model="WD7500AARX">WCACHE_OFF</disk>
       <disk model="WD30EZRSDTL">WCACHE_OFF</disk>
       <disk model="WD25EZRSDTL">WCACHE_OFF</disk>
       <disk model="WD20EARS">WCACHE_OFF</disk>
       <disk model="WD15EARS">WCACHE_OFF</disk>
       <disk model="WD10EARS">WCACHE_OFF</disk>
       <disk model="WD8000AARS">WCACHE_OFF</disk>
       <disk model="WD7500AARS">WCACHE_OFF</disk>
       <disk model="WD6400AARS">WCACHE_OFF</disk>
       <disk model="WD20EADS">WCACHE_OFF</disk>
       <disk model="WD15EADS">WCACHE_OFF</disk>
       <disk model="WD10EADS">WCACHE_OFF</disk>
       <disk model="WD8088AADS">WCACHE_OFF</disk>
       <disk model="WD7500AADS">WCACHE_OFF</disk>
       <disk model="WD6400AADS">WCACHE_OFF</disk>
       <disk model="WD5000AADS">WCACHE_OFF</disk>
       <disk model="WD3200AACS">WCACHE_OFF</disk>
       <disk model="ST32000542AS">WCACHE_OFF</disk>
       <disk model="ST31500541AS">WCACHE_OFF</disk>
       <disk model="ST31000520AS">WCACHE_OFF</disk>
       <disk model="ST3500412AS">WCACHE_OFF</disk>

[sfernandezl]

Several weeks ago, I bought two Western Digital Caviar Green 3TB SATA3, looking for large buffer to increase performance. Now, updating to version 4.0 boom. not supported. Large amount of money directly to waste.

I'm very frustrated with this move from Synology. So you create a HD support list and forget to include this little "feature"? no support of write cache for a NAS?. there was no problem if you could roll back to 3.2 if you want to avoid the lack of performance, but no, you can't. this is ridiculous, honestly.

Sure as hell my recomendation to my clients will be to avoid the companies who introduce "releases" like this one. really bad move. But hey, it's just one disappointed customer, isnt he?

[ewkid]

This is horrible news. Good thing there is nothing in this upgrade that I want. I don't want to lose performance on my green drives so I'm sticking with 3.2.

[jekyll]

Found this in var/log/msgs relating to drive cache...

Code: Select all

Mar 11 16:07:54 scemd: disk/disk_config_all.c:62 apply /usr/syno/bin/syno_disk_ctl --wcache-off  /dev/sda 1>/dev/null 2>&1
Mar 11 16:07:54 syno_disk_ctl: syno_disk_ctl.c:45 open /sys/block/sda/device/syno_deep_sleep_support fail
Mar 11 16:07:54 scemd: [/usr/syno/bin/syno_disk_ctl --wcache-off  /dev/sda 1>/dev/null 2>&1], error (255)
Mar 11 16:07:54 scemd: disk/disk_config_all.c:62 apply /usr/syno/bin/syno_disk_ctl --wcache-off  /dev/sdb 1>/dev/null 2>&1
Mar 11 16:07:54 syno_disk_ctl: syno_disk_ctl.c:45 open /sys/block/sdb/device/syno_deep_sleep_support fail
Mar 11 16:07:54 scemd: [/usr/syno/bin/syno_disk_ctl --wcache-off  /dev/sdb 1>/dev/null 2>&1], error (255)
Mar 11 16:07:54 scemd: disk/disk_config_all.c:62 apply /usr/syno/bin/syno_disk_ctl --wcache-off  /dev/sdc 1>/dev/null 2>&1
Mar 11 16:07:55 syno_disk_ctl: syno_disk_ctl.c:45 open /sys/block/sdc/device/syno_deep_sleep_support fail
Mar 11 16:07:55 scemd: [/usr/syno/bin/syno_disk_ctl --wcache-off  /dev/sdc 1>/dev/null 2>&1], error (255)
Mar 11 16:07:55 scemd: disk/disk_config_all.c:62 apply /usr/syno/bin/syno_disk_ctl --wcache-off  /dev/sdd 1>/dev/null 2>&1
Mar 11 16:07:55 syno_disk_ctl: syno_disk_ctl.c:45 open /sys/block/sdd/device/syno_deep_sleep_support fail
Mar 11 16:07:55 scemd: [/usr/syno/bin/syno_disk_ctl --wcache-off  /dev/sdd 1>/dev/null 2>&1], error (255)

Also my logs show a massive amount of errors, from USB devices (connected devices APC 1500 RM & Synology remote) below is just one of many!

Code: Select all

Mar 11 16:08:00 hotplugd: ##### ACTION:add
Mar 11 16:08:00 hotplugd: DEVICE:/proc/bus/usb/002/001
Mar 11 16:08:00 hotplugd: DEVPATH:usb2
Mar 11 16:08:00 hotplugd: SUBSYSTEM:usb
Mar 11 16:08:00 hotplugd: hotplugd.c:763 Fail to recognize usb DEVNAME=, DEVPATH=usb2
Mar 11 16:08:00 hotplugd: hotplugd.c:1289 failed to setup usb config, (DEVPATH, DEVICE, ACTION, DEVNAME, INTERFACE)=(usb2, /proc/bus/usb/002/001, add, , )
Mar 11 16:08:00 hotplugd: hotplugd.c:988 failed get usb status from usb config, DEVICE=/proc/bus/usb/002/001, usbid=1d6b:0001:0000:00:0e.0

And finally a lot of these

Code: Select all

Mar 11 16:08:16 httpd: threadproc/unix/proc.c:145 Failed to create shared memory. key[25521816], size[152992], [No such file or directory]
Mar 11 16:08:17 httpd: threadproc/unix/proc.c:145 Failed to create shared memory. key[25521816], size[152992], [No such file or directory]
Mar 11 16:08:18 httpd: threadproc/unix/proc.c:145 Failed to create shared memory. key[25521816], size[152992], [No such file or directory]

Code: Select all

Mar 11 16:07:53 root: cstn: [error]: [ShareSet]: Package is not enabled!
Mar 11 16:07:54 root: cstn: [error]: [ShareSet]: package is not enabled!
Mar 11 16:07:54 root: cstn: [error]: [ShareSet]: Package is not enabled!
Mar 11 16:07:54 root: cstn: [error]: [ShareSet]: package is not enabled!
Mar 11 16:07:54 root: cstn: [error]: [ShareSet]: Package is not enabled!
Mar 11 16:07:54 root: cstn: [error]: [ShareSet]: package is not enabled!

[tcat60]

I've been reading this topic - waiting on a statement from Synology. I'm confused by the code in the DSM upgrade that would apparently turn off write caching on some WD drives. My DS212 runs two WD20EARx.
Although this unit is primarily a media server, it also writes several daily large backups. The DS212 is running DSM version 3.2-1955. Here is why I am confused and have held off upgrade DSM - in the 3.2 version the explanation for enabling writing is to improve I/O performance (Storage Manager,Hdd management).

Reading the comments about the upgrade, it seems I am not alone in waiting for an explanation concerning this change in DSM. Could someone please spell out how this impacts perrormance of these drives? Almost everything I've read so far suggests write caching should be on.
Thank you,
tc

[synGuy]

Can you paste the disk section from /usr/syno/etc.defaults/scemd.xml. It appears that the 4.0 firmware has different content for this file for the different models of DS...which if true will be very strange indeed.

I checked scemd.xml on my 1511+ and it's the same as everyone elses (my drives listed as WCACHE_OFF), however dmesg shows all my write caches as enabled.

Odd

[microkid]

Answer from Synology support:

This feature is not available for certain specified hard drive models as it's a new designed behavior in DSM4.0 to enhance compatibility. Moreover, according to our internal system testing, this option will not have noticeable effect on disk performance so you can keep using the disks without additional concern.

Well, this doesn't explain anything. What has changed in DSM4.0 that requires this compatibility? And why can the cache safely be enabled on dozens of other drives and not on these? I haven't flashed to 4.0 but when I do, I would like to edit this file so I can enable the 64MB cache on my WD20EARS drives.

[joshuah]

So user "tjkaz" on page 2 says he has WD20EARS with write cache enabled on a DS211J, but other people with the same drives but on different DS ( more powerfull ) have it disabled!! Where is the logic in that? Especially when you consider that this worked well for everyone in 3.2, now when I checked my dmesg I see a whole bunch of new errors like that:


[359786.280000] ata2.00: status: { DRDY }
[359786.280000] ata2.00: failed command: WRITE FPDMA QUEUED
[359786.290000] ata2.00: cmd 61/10:08:e8:9b:4c/00:00:33:00:00/40 tag 1 ncq 8192 out
[359786.290000] res 40/00:1c:98:db:29/00:00:3c:00:00/40 Emask 0x10 (ATA bus error)

That never showed before in 3.2...and I don't buy that statement that there is no performance impact. The cache is there for a reason, and is not some fancy statement. Manufacturers put more cache because it helps in performance, and ESPECIALLY in writes. What compatibility is synology talking about?!?!? HW is the same, Drives are the same...and even the Linux Kernel is the same if my memory serves me well... It is in the SUPPORTED LIST of hard drives!!!

[Blueyed1]

A friend of mine just got at DS212J with a 2TB WD WC20EARS-00J99B0, there is no check mark in cache for this drive.

/Blueyed1

[sanderdw]

So here's the additional response from Synology:

It was working on DSM3.2, however for disk compatibility mechanism concern, we disable some of the write cache function of HDD like green Caviar HDD.

Based on our developer response, it should not affect much of the performance if you are using transferring file to/from your DS via your network.

Sorry , just to add additional note.

Disk compatibility mechanism is implemented in DSM4.0, that is why DSM3.2 is still working.

[joshuah]

If you do ONLY read or ONLY write from/to the NAS there might not be impact, but if you do both concurrently i.e downloading torrent file ( or 10 ) and streaming from the NAS, there should be an impact and not a small one. Why?
Well if you do a read the data gets streamed from the spindles, but if you do a write to the drive during that streaming the data can't go in the drive cache so it's stored in FileSystem cache first( no write cache ), and it has to be written immediately to the spindel, the same spindle where it's doing the reads, and that will hit the performance of the drive. That's why I assume they put so much cache in those green drives, because they are slow ( only 5400 rpm or so ) and the cache acts like a buffer, so when enough data is in the buffer it gets flushed in big chunks to the drive ( or read by the drive to the cache ) by the FW, since it knows better the chunks it will be best to offload data to the drive. If the drives were 7200 rpm or 10k they can do a lot more IO/s and they can serve those IOs much faster. Those are just my thoughts, as a layman.

[Hardcore]

Well all I can see is a lot of hearsay, from both sides.

Lets see some hard data with graphs if possible:

3.2 with the cache enabled
3.2 disabled

4.0 enabled
4.0 disabled
Once we have a clear testing methodology instead of all this "well it will definitely impact because"... ABC

Then there will be no disputing the facts either way, because if there is an impact but Synology provide graphs showing there is none, any discussion is going to be on very unstable ground


*ABC = insert bul*sh*t, unqualified statement here.

[patters]

I have a DS111 with a WD20EARS, and my write cache is enabled (and ticked in the GUI), despite it being in that list:

Code: Select all

[    3.370000] ata1: SATA link up 3.0 Gbps (SStatus 123 SControl F300)
[    3.410000] ata1.00: ATA-8: WDC WD20EARS-00MVWB0, 51.0AB51, max UDMA/133
[    3.410000] ata1.00: 3907029168 sectors, multi 0: LBA48 NCQ (depth 31/32)
[    3.460000] ata1.00: configured for UDMA/133
[    3.460000] scsi 0:0:0:0: Direct-Access     WDC      WD20EARS-00MVWB0 51.0 PQ: 0 ANSI: 5
[    3.470000] sd 0:0:0:0: [sda] 3907029168 512-byte logical blocks: (2.00 TB/1.81 TiB)
[    3.480000] sd 0:0:0:0: [sda] Write Protect is off
[    3.480000] sd 0:0:0:0: [sda] Mode Sense: 00 3a 00 00
[    3.480000] sd 0:0:0:0: [sda] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA

However, I had disabled my drive's idle timer before I fitted it into my syno. Could this be why mine works? I seem to have the same /usr/syno/etc.defaults/scemd.xml file as you guys (my drive is in the list as WCACHE_OFF).

[joshuah]

Patters, can you check for messages about write cache disabled towards the end of dmesg? In the beginning of the boot process I have it enabled too, but toward the end it's being disabled...