Bad Blocks?

                   uptime: 1m38s
                  version: 6.36 (stable)
               build-time: Jul/20/2016 14:09:10
              free-memory: 101.9MiB
             total-memory: 128.0MiB
                      cpu: MIPS 74Kc V4.12
                cpu-count: 1
            cpu-frequency: 600MHz
                 cpu-load: 2%
           free-hdd-space: 108.4MiB
          total-hdd-space: 128.0MiB
  write-sect-since-reboot: 217
         write-sect-total: 518172
               bad-blocks: 0.4%
        architecture-name: mipsbe
               board-name: RB2011UiAS-2HnD
                 platform: MikroTik

what’s wrong with my router?

The internal storage is failing. 0,4% of its blocks are marked bad. This is what happens when the system cannot use these sectors anymore.

It’s quite safe if the value doesn’t rise. Use external flash for files operations and logging instead the internal to protect it from the next damages.

Just a thought: It wrote only 518172 sectors. Isn’t it a little… to little to start failing?

Up to 0.5% can be marked as bad from the factory already. This is normal and all flash manufacturers mark some sectors as bad, to avoid any software using them. The only issue is if you see the bad % increase every day. If it stays, it is normal.

I thought his bad blocks were increasing. :smiley:

But it got me worried. I assumed each sector would be 512 bytes, the write endurance would be something around 3k and the flash controller would do wear leveling. Am I right? Can you give us some ballpark number? Knowing these numbers would be easy to estimate tradeoffs about the flash use.

You can google the NAND chip model in question, the manufactuers datasheet contains the needed info

But where do I find this info? Unfortunately I can’t open the device: is is sealed by the reseller, and I would loose the warranty.

Thanks,

Depends on the model that you have. For example http://www1.futureelectronics.com/doc/SAMSUNG/K9F1G08U0C-PCB000.pdf
See section about bad blocks. As you can see, some bad blocks are OK from the chip manufacturer.

Yes, they are. Thanks for the info. I will try to find my flash number.

Adding some kind of new bad blocks since last reboot and new bad blocks since initial boot/reset config may be a nice way to detect failing NAND, and it seems like next to zero work.

One of our CSR226-24 is also showing quite a high amount of writes - luckily no bad blocks yet

[b]Uptime		05:27:35[/b]
...
[b]Sector Writes Since Reboot		1 015[/b]
Total Sector Writes		1 184 744
Bad Blocks		0.0 %

The device is now in operation for approx one year and I do really wonder why it got around 200 writes per hour as it’s just serving as a switch and running a rarely used ovpn-server for maintenance access.

Graphs are being kept in memory and so do the logs. Any idea what can be tuned to reduce the writes?! How many writes can we expect the device to endure?
We’re also having some CRS125 in operation running almost the same config and we don’t notice this sort of behaviour there…

Maybe the ros version differs. Some older versions had error causing excessive writes to flash…

Not really, the CRS226 is running the latest 6.36.2

And for the last 5 days I got almost 20k blocks written. Any idea what might be causing this?
Possibly DHCP leases??

netinstall does not solve the problem.
Can I resolder the nand chip and keep the license ?

I have three such devices and everywhere the same symptoms.
The device 2012 yr.
Firmware version does not matter.

The only thing I have installed microsd card. After tomorrow , I’ll try to remove it and do the netinstall again.

With just 100k sector writes having 91% bad blocks is not normal. Especially if you have multiple boards with the same stats as you imply.
Forget de-soldering the NAND chip. Just send it to RMA.

I’ve got boards with hundreds of thousands (or even millions) of writes without any bad blocks (or at most 0.1%).
Screenshot_1.png
Screenshot_2.png
Screenshot_3.png



[amir@MikroTikFA-RB2011] /system resource> pri
                   uptime: 2w4d4h16m58s
                  version: 6.40.4 (stable)
               build-time: Oct/02/2017 08:38:30
              free-memory: 103.7MiB
             total-memory: 128.0MiB
                      cpu: MIPS 74Kc V4.12
                cpu-count: 1
            cpu-frequency: 600MHz
                 cpu-load: 4%
           free-hdd-space: 105.5MiB
          total-hdd-space: 128.0MiB
  write-sect-since-reboot: 18402
         write-sect-total: 112080
               bad-blocks: 2.2%
        architecture-name: mipsbe
               board-name: RB2011UiAS-2HnD
                 platform: MikroTikFA

now

There is something very wrong with your units. Compare with this one.

                   uptime: 1w6d12h41m38s
                  version: 6.40.4 (stable)
               build-time: Oct/02/2017 08:38:30
              free-memory: 40.4MiB
             total-memory: 64.0MiB
                      cpu: MIPS 24Kc V7.4
                cpu-count: 1
            cpu-frequency: 650MHz
                 cpu-load: 9%
           free-hdd-space: 4692.0KiB
          total-hdd-space: 16.0MiB
  write-sect-since-reboot: 12732
         write-sect-total: 444795
               bad-blocks: 0%
        architecture-name: mipsbe
               board-name: hAP ac lite
                 platform: MikroTik

@Paternot @Cha0s @irghost

My respect, but you decided to play captain obvious ?
I have many other devices everything is fine .

At the moment I have three 493G , which needs to be fixed.The question is only about the license . How can I keep it ?


I have ended the warranty on these devices.
100k writes, it is necessary to consider, that would be reset after netinstall. He does not keep the total number of the recorded sectors. I will say that there was no alloying or web proxy.


At the current moment.

Now will be reset and downgrading to routeros version 5. Netinstall cannot install 6 version.Then upgrade to version 6 bug fix only.Will remove the sd card and will be watching.

What that’s supposed to mean?

I don’t think that’s true. I just did a netinstall on a spare old RB I have here at the office and the Total Sectors Writes was not reset.
Screenshot_2.png
Screenshot_3.png
That’s the expected (and proper) behavior IMHO. What you suggested is like buying a used car (RB) and with just a change of tires (netinstall) you get a reset odometer wiping the 300.000kms (millions of sector writes) you did while you had it, then selling it like new to some unaware guy…

Here’s another board I just noticed today. 8.8 million sector writes. 0.0% bad blocks.
Screenshot_1.png
All I am saying is, your board is pretty much dead (or soon will totally be) with 97% bad blocks. Just replace it if it’s out of warranty.