7 ms·

I used to build my own stuff, but now I buy Synology. It's really convenient, has great UI and support, a bunch of different packages you can install, etc. Al

I used to build my own stuff, but now I buy Synology. It's really convenient, has great UI and support, a bunch of different packages you can install, etc. Although I ran into a problem a couple of months after setting it up where the NAS became so slow, it was unusable. I saw that my IOWait was 100%, so I figured it was a disk, but nothing indicated a problem. Eventually I was able to get to a log that showed that one of the drives had some weird error messages, so I bought a new drive from Amazon that came the next day, pulled the drive, and replaced it and instantly everything was okay again. I would have expected something on Synology's status programs to show a problem, but they were all green, so that was annoying.

voltagex_ 11y ago
The Linux kernel can do some really strange things when a drive fails. I had an external drive fail to the point where if I wrote to it, commands like ls would fail and be unkillable until I rebooted the system. I wonder if that's similar to what you were experiencing, masked by the shiny UI
- rincebrain 11y ago
  It's not limited to Linux - the problem is that, while you can go around resetting increasingly large parts of the IO stack, lots of devices in the path may or may not actually respond to "off" or "reset" when they're in a bad state, so a drive being in a bad state can cause the entire IO path it's in to go south for the winter if nobody is smart enough in the path. This is one of the actual reasons that enterprise drives can be better - almost all of them support an equivalent of TLER (time-limited error recovery), which is basically a programmable timeout on read/write errors. Most parts of the IO stack, hardware and software, on every platform I've used deal a lot better with explicit errors than a device that hasn't appeared to vanish but is acting like a black hole.
Jedd 11y ago
> It's really convenient, ... > ... would have expected something on Synology's status > programs to show a problem, but they were all green, > so that was annoying. So perhaps not all that convenient. I went down the QNap path a couple of years ago, on the assumption it'd be robust, well patched, more convenient, and basic functionality would work / be tested. I was wrong on most counts, and am now basically scared off these low-end SOHO systems entirely.
- pyre 11y ago
  It's convenient most of the time, but sometimes building your own stuff is inconvenient most of the time. It's all about trade-offs. For example, a while back I had to track down an issue on a server with Nagios constantly triggering memory low errors. Turns own that when you stat a lot of files, XFS will load all of that file information into cache in the kernel's slab. This causes some RAM stats to be skewed. This RAM will be instantly freed if needed, but some system stats count it towards "RAM usage." My question is: How does one incorporate all such edge-cases into something that is user-friendly? How many system admins even know how to track down such an issue (in the absence of a flashy UI)? Unless you want to vertically integrate the system from hardware => kernel => UI, then you're bound to have these sorts of issues where everyone is only paying attention with what's happening in their own little area of concern and multiple decoupled systems can end up interacting poorly (even when their local actions make sense to them).
  - laumars 11y ago
    Re cached memory, isn't that how all file systems work across all the major operating systems? The kernel space will cache files to reduce disk IO but will immediately free that memory if a requesting application allocated more memory that's what's available in free memory alone. Which is why file servers perform best with ooodles of RAM and why one should always look at both the free + cached memory when calculating available memory on more general purpose systems. Or was your NAS behaving subtly different from the standard above? (eg how ZFS cache doesn't appear as FS cache in ZoL despite freeing in much the same way as the above)
    - pyre 11y ago
      This is a specific XFS behaviour. In this case, the slab was using up 4GB of RAM on a 8GB machine, and all of the normal tools were telling us that we were running out of free space in RAM. No such problems on other file systems.
  - e40 11y ago
    What they offer is a good form factor with lower power. What would you suggest for a home brew, that would match size and power consumption?
nugget 11y ago
Out of curiosity what RAID setup would you use for replaceable media (e.g. blu ray movies)? Is there a synology solution for checksums to pevent bitrot which ZFS advocates talk about a lot?
- pfarnsworth 11y ago
  You can run periodic scans to look for bitrot. I think this is the standard way of handling this, my ReadyNAS before this had the same functionality.
  - usefulcat 11y ago
    How can it look for bitrot without any checksum info?
    - rakoo 11y ago
      Instead of running some homemade script, you better use something like SnapRaid (http://www.snapraid.it/ http://www.snapraid.it/)
    - pfarnsworth 11y ago
      RAID has redundant information so they use that.
      - Moru 11y ago
        Depending on what RAID you are running, you will only know you have bitrot, you can't fix it since you don't know what harddrive has the correct information.
      - cmurf 11y ago
        This doesn't unambiguously detect mismatches though. If it's a mirror, you have two copies of data, which one is bad? All RAID 1 knows is that they're different. If it's RAID456, again, all it knows is that there's a mismatch, it doesn't know if the data strips are wrong, or if the parity strips are wrong. The way ZFS and Btrfs deal with this is data is checksummed, and the fs metadata which includes the parity strips, are checksummed. So there's a completely independent way of knowing what's incorrect, rather than merely that there's a difference.
okigan 11y ago
Also got a synology -- pretty UI, but keep running into weird limitation (ex. only two timed snapshot, inability to get standard linux tools, ex rsnapshot, DLNA slow to list files, PhotoStation unusably slow, fs weird and does not logically match to what is presented in UI) And built a multi VM box (with VM in VM support): https://pcpartpicker.com/user/okigan/saved/#view=nmQD4D https://pcpartpicker.com/user/okigan/saved/#view=nmQD4D Btw, pcpartpicker was useful during my built, Node 804 was a case I considered, but wanted to have an External 5.25.
- unholythree 11y ago
  I like pcpartpicker quite a bit especially their $/GB search feature for hard drives. I do wish they would include some basic server hardware for enthusiasts though, big ECC dimms are getting affordable and it would be nice to do some price comparison.
jeffcox 11y ago
You can buy a FreeNAS box built by the same company that runs the project on Amazon. A bit pricey, but it's been perfect for me for over a year.
cthalupa 11y ago
I cannot recommend prosumer level synology gear after having had nightmare after nightmare with it. The chips not used in their pro line are just not powerful enough for rebuilding large arrays in a reasonable amount of time. RAID6 rebuilds of a 12 4tb drive array take nearly a full week on my DS1812+ I'm in the process of moving everything to a OmniOS server and will not look back.
emsy 11y ago
The prices for the larger ones are ridiculous though. 800$ for a 5-bay NAS without HDDs is a lot. You could easily build 2 or more NASes with the same specs for that money. Also, having used a Synology I think it's a niche product. It's too powerful for people who simply want to put files and backups on their network. But for people who like to tinker and control, it's too restricted. You can basically only install software that has been ported to Synology. Getting it to do automated tasks the way you want to can be annoyingly complicated. I would recommend it to an advanced customer, who needs features like Owncloud or VNP without having the burden to maintain a system.
- jdhawk 11y ago
  xpenology lets you run synology on commodity hardware - I have a 212j, and run xpenology on an old Optiplex I had lying around for Plex with Transcoding. pretty neat
- briffle 11y ago
  I was playing with the parts in this article, and it was up near $2500 from amazon. $250MB motherboard, 4x$60 8GB dimms, etc.
  - emsy 11y ago
    I was talking about the 5bay Synology model.
- soylentcola 11y ago
  For me it was a matter of wanting basic network backup and storage as well as DVR/storage for a couple of IP security cams I have around the house. I know I could have set up something from scratch but instead I just went with the Synology for ease of setup, relatively affordable price (didn't have enough of the right parts laying around for a franken-build), and as you say, less burden of maintenance. I wouldn't necessarily want to depend on it for something "mission critical" but it's a convenient solution along with online backup for media storage and security cam management. My only regret is not buying one with a more powerful processor since this one isn't really capable of transcoding media. Instead, I have all of my home media backed up to the Synology and run a Plex server on my primary desktop. The Plex server reads the media from the NAS and can then send to Chromecast or be accessed from Plex or Kodi running on my Android TV.

voltagex_ 11y ago

The Linux kernel can do some really strange things when a drive fails. I had an external drive fail to the point where if I wrote to it, commands like ls would fail and be unkillable until I rebooted the system. I wonder if that's similar to what you were experiencing, masked by the shiny UI

rincebrain 11y ago
It's not limited to Linux - the problem is that, while you can go around resetting increasingly large parts of the IO stack, lots of devices in the path may or may not actually respond to "off" or "reset" when they're in a bad state, so a drive being in a bad state can cause the entire IO path it's in to go south for the winter if nobody is smart enough in the path. This is one of the actual reasons that enterprise drives can be better - almost all of them support an equivalent of TLER (time-limited error recovery), which is basically a programmable timeout on read/write errors. Most parts of the IO stack, hardware and software, on every platform I've used deal a lot better with explicit errors than a device that hasn't appeared to vanish but is acting like a black hole.

Jedd 11y ago

> It's really convenient, ... > ... would have expected something on Synology's status > programs to show a problem, but they were all green, > so that was annoying. So perhaps not all that convenient. I went down the QNap path a couple of years ago, on the assumption it'd be robust, well patched, more convenient, and basic functionality would work / be tested. I was wrong on most counts, and am now basically scared off these low-end SOHO systems entirely.

pyre 11y ago
It's convenient most of the time, but sometimes building your own stuff is inconvenient most of the time. It's all about trade-offs. For example, a while back I had to track down an issue on a server with Nagios constantly triggering memory low errors. Turns own that when you stat a lot of files, XFS will load all of that file information into cache in the kernel's slab. This causes some RAM stats to be skewed. This RAM will be instantly freed if needed, but some system stats count it towards "RAM usage." My question is: How does one incorporate all such edge-cases into something that is user-friendly? How many system admins even know how to track down such an issue (in the absence of a flashy UI)? Unless you want to vertically integrate the system from hardware => kernel => UI, then you're bound to have these sorts of issues where everyone is only paying attention with what's happening in their own little area of concern and multiple decoupled systems can end up interacting poorly (even when their local actions make sense to them).
- laumars 11y ago
  Re cached memory, isn't that how all file systems work across all the major operating systems? The kernel space will cache files to reduce disk IO but will immediately free that memory if a requesting application allocated more memory that's what's available in free memory alone. Which is why file servers perform best with ooodles of RAM and why one should always look at both the free + cached memory when calculating available memory on more general purpose systems. Or was your NAS behaving subtly different from the standard above? (eg how ZFS cache doesn't appear as FS cache in ZoL despite freeing in much the same way as the above)
  - pyre 11y ago
    This is a specific XFS behaviour. In this case, the slab was using up 4GB of RAM on a 8GB machine, and all of the normal tools were telling us that we were running out of free space in RAM. No such problems on other file systems.
- e40 11y ago
  What they offer is a good form factor with lower power. What would you suggest for a home brew, that would match size and power consumption?

nugget 11y ago

Out of curiosity what RAID setup would you use for replaceable media (e.g. blu ray movies)? Is there a synology solution for checksums to pevent bitrot which ZFS advocates talk about a lot?

pfarnsworth 11y ago
You can run periodic scans to look for bitrot. I think this is the standard way of handling this, my ReadyNAS before this had the same functionality.
- usefulcat 11y ago
  How can it look for bitrot without any checksum info?
  - rakoo 11y ago
    Instead of running some homemade script, you better use something like SnapRaid (http://www.snapraid.it/ http://www.snapraid.it/)
  - pfarnsworth 11y ago
    RAID has redundant information so they use that.
    - Moru 11y ago
      Depending on what RAID you are running, you will only know you have bitrot, you can't fix it since you don't know what harddrive has the correct information.
    - cmurf 11y ago
      This doesn't unambiguously detect mismatches though. If it's a mirror, you have two copies of data, which one is bad? All RAID 1 knows is that they're different. If it's RAID456, again, all it knows is that there's a mismatch, it doesn't know if the data strips are wrong, or if the parity strips are wrong. The way ZFS and Btrfs deal with this is data is checksummed, and the fs metadata which includes the parity strips, are checksummed. So there's a completely independent way of knowing what's incorrect, rather than merely that there's a difference.

okigan 11y ago

Also got a synology -- pretty UI, but keep running into weird limitation (ex. only two timed snapshot, inability to get standard linux tools, ex rsnapshot, DLNA slow to list files, PhotoStation unusably slow, fs weird and does not logically match to what is presented in UI) And built a multi VM box (with VM in VM support): https://pcpartpicker.com/user/okigan/saved/#view=nmQD4D https://pcpartpicker.com/user/okigan/saved/#view=nmQD4D Btw, pcpartpicker was useful during my built, Node 804 was a case I considered, but wanted to have an External 5.25.

unholythree 11y ago
I like pcpartpicker quite a bit especially their $/GB search feature for hard drives. I do wish they would include some basic server hardware for enthusiasts though, big ECC dimms are getting affordable and it would be nice to do some price comparison.

jeffcox 11y ago

You can buy a FreeNAS box built by the same company that runs the project on Amazon. A bit pricey, but it's been perfect for me for over a year.

cthalupa 11y ago

I cannot recommend prosumer level synology gear after having had nightmare after nightmare with it. The chips not used in their pro line are just not powerful enough for rebuilding large arrays in a reasonable amount of time. RAID6 rebuilds of a 12 4tb drive array take nearly a full week on my DS1812+ I'm in the process of moving everything to a OmniOS server and will not look back.

emsy 11y ago

The prices for the larger ones are ridiculous though. 800$ for a 5-bay NAS without HDDs is a lot. You could easily build 2 or more NASes with the same specs for that money. Also, having used a Synology I think it's a niche product. It's too powerful for people who simply want to put files and backups on their network. But for people who like to tinker and control, it's too restricted. You can basically only install software that has been ported to Synology. Getting it to do automated tasks the way you want to can be annoyingly complicated. I would recommend it to an advanced customer, who needs features like Owncloud or VNP without having the burden to maintain a system.

jdhawk 11y ago
xpenology lets you run synology on commodity hardware - I have a 212j, and run xpenology on an old Optiplex I had lying around for Plex with Transcoding. pretty neat
briffle 11y ago
I was playing with the parts in this article, and it was up near $2500 from amazon. $250MB motherboard, 4x$60 8GB dimms, etc.
- emsy 11y ago
  I was talking about the 5bay Synology model.
soylentcola 11y ago
For me it was a matter of wanting basic network backup and storage as well as DVR/storage for a couple of IP security cams I have around the house. I know I could have set up something from scratch but instead I just went with the Synology for ease of setup, relatively affordable price (didn't have enough of the right parts laying around for a franken-build), and as you say, less burden of maintenance. I wouldn't necessarily want to depend on it for something "mission critical" but it's a convenient solution along with online backup for media storage and security cam management. My only regret is not buying one with a more powerful processor since this one isn't really capable of transcoding media. Instead, I have all of my home media backed up to the Synology and run a Plex server on my primary desktop. The Plex server reads the media from the NAS and can then send to Chromecast or be accessed from Plex or Kodi running on my Android TV.