How many (aproximately) writes to flash will kill it?
I have:
[admin@TESTY] > /system resource print
uptime: 4d20h48m45s
version: "2.9.25"
free-memory: 15744kB
total-memory: 30440kB
cpu: "MIPS 4Kc V0.10"
cpu-frequency: 264MHz
cpu-load: 0
free-hdd-space: 93816kB
total-hdd-space: 126976kB
write-sect-since-reboot: 34491
write-sect-total: 103900
bad-blocks: 1
Should I worry about last two numbers? Is 100.000 written sector really big value? Should I consider changing config & preparing new flash with backup?
Gregor
Code: Select all
write-sect-since-reboot: 129728
write-sect-total: 3556164
True? Server or client?This is an older Sandisk...still chiggin' away. Note that if you use NTP, writes will increase much faster.Code: Select allwrite-sect-since-reboot: 129728 write-sect-total: 3556164
Mike
Code: Select all
write-sect-since-reboot: 3915872
write-sect-total: 4417528
Supposedly you get about 300,000 writes per sector and the flash itself manages the wear leveling, so I don't know if its worth worrying about anymore.
Sam
In my experiences, it's been with the NTP Client. I actually don't use the client anymore because of this (and my paranoia about the flash failing) although it is still running on this particular router. The uptime of that router I posted stats on is only ~4 days, and there's been 129,000 writes to it already - the ntp client is running on this. DHCP server "store leases to disk" is set to never. There's really not much else running on this that would cause this many writes. It's a hotspot gateway. There is no logging to disk either.
In contrast, here's another with basically the same config (this network is actually busier), but NTP client is not running:
Only 1000 writes total in 2 days vs. 129,000 over 4 days on the other one.
As for the PQI flash cards...good cards, but man these killed us awhile back as they don't work on a Soekris4801
In contrast, here's another with basically the same config (this network is actually busier), but NTP client is not running:
Code: Select all
uptime: 2d1h10m34s
version: "2.9.24"
write-sect-since-reboot: 988
write-sect-total: 617168As for the PQI flash cards...good cards, but man these killed us awhile back as they don't work on a Soekris4801
I use ntp client and I have big number of writes. Mikrotik? Can you do something with ntp client?This is an older Sandisk...still chiggin' away. Note that if you use NTP, writes will increase much faster.Code: Select allwrite-sect-since-reboot: 129728 write-sect-total: 3556164
Have never had a problem with any of my flash modules
That one had been around since 2.7.x
Code: Select all
uptime: 3w2d3h13m3s
version: "2.9.24"
write-sect-since-reboot: 2978210
write-sect-total: 13400762
This is good news for a lot of people I think. What brand (or chip) of flash card is this?That one I can't remember exactly, it's one of the direct IDE plugins into a standard motherboard, acting as my primary router with 9 network interfaces. All of which are graphed. Been thinking about putting a second disk on chip module for storing graphs and logs. Haven't checked into it fully, but I could have sworn that I heard it was possible. As you can see I get some use out of my equipment. 
And so far (Knock on wood) not a single problem. For all my compact flash modules I have been sticking with SanDisk, for testing I have been using Verbatim (got them cheap). A little off topic, but for my web proxy drives I only use Seagate drives. Everything always well tested even before thinking about loading anything on them.
And so far (Knock on wood) not a single problem. For all my compact flash modules I have been sticking with SanDisk, for testing I have been using Verbatim (got them cheap). A little off topic, but for my web proxy drives I only use Seagate drives. Everything always well tested even before thinking about loading anything on them.
80GB HDD for MT :)
I have not a compact flash drive, and my mikrotik is installed on.... 80GB Seagate drive 
offtopic
like my avatarsometimes those die sooner
This is (or was?) 540MB Seagate....http://www.embeddedintel.com/display.php?article=53
Nice short read on flash.
As stated in another thread, my edge router is running the same module from WISPCON II in '03 with no issues with 31,412,898 total writes. I think that number is much bigger; there was an upgrade way back where the total got reset. Of course the PC based MT doesn't show the bad block count like the RBs.
Thanks for the comment about NTP and echo logging. If we simply turn off any echo the the writes go away? We can't see console anyway...
Mike
Nice short read on flash.
As stated in another thread, my edge router is running the same module from WISPCON II in '03 with no issues with 31,412,898 total writes. I think that number is much bigger; there was an upgrade way back where the total got reset. Of course the PC based MT doesn't show the bad block count like the RBs.
Thanks for the comment about NTP and echo logging. If we simply turn off any echo the the writes go away? We can't see console anyway...
Mike
Action=echo I have only for critical logs. I have seen that ntp client increases number of writes - please check it (in your lab).NTP client isn't writing anything to the drive.
"/system logging action=echo" is increasing the sector writes.
Next version will have more optimized management so it isn't writing so much.
Gregor
could you send the support output file form this router to support@mikrotik.com?2.9.26
32 hours
7376 sector writes since reboot. Seems high as I don't know what ROS would need to write.
Minimal config: Routing and NTP client. Normal drops in firewall. No web-proxy, DHCP, hotspot or nat. It's just a router sitting at my fiber edge.
On the way!could you send the support output file form this router to support@mikrotik.com?2.9.26
32 hours
7376 sector writes since reboot. Seems high as I don't know what ROS would need to write.
Minimal config: Routing and NTP client. Normal drops in firewall. No web-proxy, DHCP, hotspot or nat. It's just a router sitting at my fiber edge.
Edit:
I pulled a RB112 from the field; no difference in writes with NTP client on or off.
Hi,wow, very nice examples, people. thanks!
Here is another one
Code: Select all
uptime: 44d5h15m20s
free-memory: 105436 kB
total-memory: 127584 kB
cpu: AMD
cpu-frequency: 1000 MHz
cpu-load: 1
free-hdd-space: 30969 kB
total-hdd-space: 61972 kB
write-sect-since-reboot: 126530
write-sect-total: 3295904
I have a RB500 that appears kinda high -
[magicwisp@MagicWISPMT2] system resource> print
uptime: 4d10h29m1s
version: "2.9.26"
free-memory: 13844kB
total-memory: 30440kB
model: "RouterBOARD 500"
cpu: "MIPS 4Kc V0.10"
cpu-frequency: 333MHz
cpu-load: 36
free-hdd-space: 5648kB
total-hdd-space: 61440kB
write-sect-since-reboot: 886727
write-sect-total: 27476167
bad-blocks: 1960
[magicwisp@MagicWISPMT2] system resource> print
uptime: 4d10h29m1s
version: "2.9.26"
free-memory: 13844kB
total-memory: 30440kB
model: "RouterBOARD 500"
cpu: "MIPS 4Kc V0.10"
cpu-frequency: 333MHz
cpu-load: 36
free-hdd-space: 5648kB
total-hdd-space: 61440kB
write-sect-since-reboot: 886727
write-sect-total: 27476167
bad-blocks: 1960
Hi,wow, very nice examples, people. thanks!
Here is another oneThis flash works since 2.7.14 without any problems.Code: Select alluptime: 44d5h15m20s free-memory: 105436 kB total-memory: 127584 kB cpu: AMD cpu-frequency: 1000 MHz cpu-load: 1 free-hdd-space: 30969 kB total-hdd-space: 61972 kB write-sect-since-reboot: 126530 write-sect-total: 3295904
[] > sys reso prin
uptime: 1w1d4h25m40s
version: "2.9.27"
free-memory: 450416kB
total-memory: 516732kB
cpu: "Intel(R)"
cpu-frequency: 3192MHz
cpu-load: 18
free-hdd-space: 25030kB
total-hdd-space: 61484kB
write-sect-since-reboot: 1167028
write-sect-total: 38686680
[] >
Just graphing on several interfaces. Is there a way to install a conventional HDD just for storing graphs?
Best,
Brad