search

openwrt / packages

Community maintained packages for OpenWrt. Documentation for submitting pull requests is in CONTRIBUTING.md
GNU General Public License v2.0
3.98k stars 3.47k forks source link

luci_statistics not flushing it's cache on router reboot #2604

Closed kylejvrsa closed 7 years ago

kylejvrsa commented 8 years ago

I have a usb flash drive connected to my router. I use this drive for storing all luci_statistics (rrd graphs), because this is a USB flash drive I want to limit the writes to it as they don't last long when writing allot, so I flush the collectd cache every 10mins, when I need to restart luci_statistics (/etc/luci_statistcis restart) it flushes the cache, however when I reboot the router it doesn't flush the cache it just reboots and I lose all the gathered data in the 10min.

Router: WRT1900AC Firmware: OpenWRT 15.05

ghost commented 8 years ago

Silly question, but why do you think this matters? Do you have some data that wear levelling doesn't work even remotely close to theory so that even with 10 years of ridiculously frequent writes on even 8 MB flash you wouldn't be fine?

Regards,

Daniel

On 16-04-15 01:31 AM, kylejvrsa wrote:

I have a usb flash drive connected to my router. I use this drive for storing all luci_statistics (rrd graphs), because this is a USB flash drive I want to limit the writes to it as they don't last long when writing allot, so I flush the collectd cache every 10mins, when I need to restart luci_statistics (/etc/luci_statistcis restart) it flushes the cache, however when I reboot the router it doesn't flush the cache it just reboots and I lose all the gathered data in the 10min.

Router: WRT1900AC Firmware: OpenWRT 15.05

— You are receiving this because you are subscribed to this thread. Reply to this email directly or view it on GitHub https://github.com/openwrt/packages/issues/2604

kylejvrsa commented 8 years ago

I just thought I'd mention this, as I was writing to the drive very frequently and the flash drive failed within 4 months so I increased the cache flush in the hope that writing to it less may increase it's lifespan. Anyway so what is your take on the cache being flushed on reboot?

Cheers!

ghost commented 8 years ago

What sort of failure for the flash drive? i.e. I've had usb flash drives not being written often fail due to crappy usb electronics that do nasty things to the electronics of the drive when the device had constant moderate cpu and higher temperatures than stock units normally operate with.

That is I'm not convinced reduced flush is actually doing to do anything for you.

Regards,

Daniel

On 16-04-15 02:01 AM, kylejvrsa wrote:

I just thought I'd mention this, as I was writing to the drive very frequently and the flash drive failed within 4 months so I increased the cache flush in the hope that writing to it may increase it's lifespan. Anyway so what is your take on the cache being flushed on reboot?

Cheers!

— You are receiving this because you commented. Reply to this email directly or view it on GitHub https://github.com/openwrt/packages/issues/2604#issuecomment-210299261

kylejvrsa commented 8 years ago

Whether I will increase the lifespan of the flash drive with increased cache flush time might help or it might not. The reason I created the issue was to find out whether the luci_statistics should flush it's cache or not. So what is your take on cache being flushed on router reboot? I would think that, that should be done even if it's just 30 seconds of cache.

ghost commented 8 years ago

Sorry just noticed what I suspect is an urban legend about flash lifespan (I've had conversations with a product manager who was convinced that the fact that even (even years ago) cheap flash has wear levelling built in, and I read the specs and did some calculations and came to conclusion that unless manufacturers are managing to pass off total fiction and not getting sued for it, that flash writes are really a non-issue except under extremely unusual circumstances).

I can't say I've exactly researched the topic, but what little searching I did only comes up with anecdotal theories that the cause of failure was excessive writes.

The concept gets bandied around so much without hard data that I know about that it smells to me like classic urban legend (albeit it's kind of like the battery memory legend - it was once true, but not with modern gear (although batteries are a little different as they're still subject to maintenance issues, about which you hear a variety of stories for what's the best way to mitigate, there seem to be a number of pet theories about that)).

Anyway for flushing cache on reboot I don't think that's practical unless you can arrange for the reboot to only be done by software (in which case the shutdown of collectd should already take care of it).

If you hit the power button there isn't a means on the vast majority of router hardware (if not all) to intercept the the button press and perform pc-like shutdown or reboot.

Regards,

Daniel

On 16-04-15 02:20 AM, kylejvrsa wrote:

Whether I will increase the lifespan of the flash drive with increased cache flush time might help or it might not. The reason I created the issue was to find out whether the luci_statistics should flush it's cache or not. So what is your take on cache being flushed on router reboot? I would think that, that should be done even if it's just 30 seconds of cache.

— You are receiving this because you commented. Reply to this email directly or view it on GitHub https://github.com/openwrt/packages/issues/2604#issuecomment-210302348

kylejvrsa commented 8 years ago

If a server is restarted, a service should be successfully stopped, in collectd's case I would assume it should flush the cache before it's shutdown right?

Cheers,

ghost commented 8 years ago

If restarted by e.g. reboot (but not reboot -f that bypasses initscripts), it ought to flush the cache because it should be shutdown cleanly.

Regards,

Daniel

On 2016-04-15 04:38 AM, kylejvrsa wrote:

If a server is restarted, a service should be successfully stopped, in collectd's case I would assume it should flush the cache before it's shutdown right?

Cheers,

— You are receiving this because you commented. Reply to this email directly or view it on GitHub https://github.com/openwrt/packages/issues/2604#issuecomment-210361747

hnyman commented 8 years ago

This might have been fixed in DD trunk by https://github.com/openwrt/packages/commit/8228cee3f34bbae17a1625fe10e9544dbef01581 (The commit added a STOP value for the collectd init script, so that a shutdown will cause the service to stop. But I haven't tested if that really works for collectd.)

kylejvrsa commented 8 years ago

@dfc-d: I use luci to reboot or reboot from terminal.

@hnyman: Cool! I can't test it as I am running 15.05, been pushing for us to get another device here so I can test trunk. Would be really cool if it is indeed fixed as that is the expected behaviour.

Thanks to both of you for responding :)

hnyman commented 8 years ago

@kylejvrsa

You can easily test it in your 15.05 router. Just check the change.... It is just one line added to the text file /etc/init.d/collectd The line STOP=10 has been added after that START line

You can edit the file by yourself https://github.com/openwrt/packages/blob/8228cee3f34bbae17a1625fe10e9544dbef01581/utils/collectd/files/collectd.init

diizzyy commented 7 years ago

@hnyman Please close this, wrong repo and 15.X is unsupported.