From: Ivan L. Jr. <iva...@gm...> - 2013-12-07 12:47:00
|
Hello, while reading the man pages I realized that there were some suggestions as to how often certain types of tests could be run. Are they arbitrary or can they be used at least as a sensible starting point? Per man pages long test can be run weekly, short test daily, automatic offline immediate could be run per vendor default choice of typically each 4 hours and so on. Well, how often should one run these tests? Any known best practices? Recommendations? Ivan |
From: Ivan L. Jr. <iva...@gm...> - 2013-12-13 07:49:16
|
So, no consensus on this or … ? On Dec 7, 2013, at 2:46 PM, Ivan Lezhnjov Jr. <iva...@gm...> wrote: > Hello, > > while reading the man pages I realized that there were some suggestions as to how often certain types of tests could be run. Are they arbitrary or can they be used at least as a sensible starting point? > > Per man pages long test can be run weekly, short test daily, automatic offline immediate could be run per vendor default choice of typically each 4 hours and so on. > > Well, how often should one run these tests? Any known best practices? Recommendations? > > Ivan |
From: Weedy <wee...@gm...> - 2013-12-15 02:58:59
|
A very long time ago I found this and use it everywhere now. I don't even remember what the options mean anymore. /dev/sda -d ata -o off -S on -a -s (O/../../(7|3)/00|S/../.././12|L/../../6/03|C/../../7/03) -m root I have a follow up question for this. I can drop the /dev/sdX and use DEVICESCAN right? How would I tell smartd to stagger the tests if I have a 4+ drive raid? On Fri, Dec 13, 2013 at 2:49 AM, Ivan Lezhnjov Jr. < iva...@gm...> wrote: > So, no consensus on this or … ? > > On Dec 7, 2013, at 2:46 PM, Ivan Lezhnjov Jr. <iva...@gm...> > wrote: > > > Hello, > > > > while reading the man pages I realized that there were some suggestions > as to how often certain types of tests could be run. Are they arbitrary or > can they be used at least as a sensible starting point? > > > > Per man pages long test can be run weekly, short test daily, automatic > offline immediate could be run per vendor default choice of typically each > 4 hours and so on. > > > > Well, how often should one run these tests? Any known best practices? > Recommendations? > > > > Ivan > > > > ------------------------------------------------------------------------------ > Rapidly troubleshoot problems before they affect your business. Most IT > organizations don't have a clear picture of how application performance > affects their revenue. With AppDynamics, you get 100% visibility into your > Java,.NET, & PHP application. Start your 15-day FREE TRIAL of AppDynamics > Pro! > http://pubads.g.doubleclick.net/gampad/clk?id=84349831&iu=/4140/ostg.clktrk > _______________________________________________ > Smartmontools-support mailing list > Sma...@li... > https://lists.sourceforge.net/lists/listinfo/smartmontools-support > |
From: Ivan L. Jr. <iva...@gm...> - 2013-12-15 06:06:16
|
On Dec 15, 2013, at 4:58 AM, Weedy <wee...@gm...> wrote: > A very long time ago I found this and use it everywhere now. I don't even remember what the options mean anymore. > /dev/sda -d ata -o off -S on -a -s (O/../../(7|3)/00|S/../.././12|L/../../6/03|C/../../7/03) -m root I read all man pages, was a bit painful haha but I can read that line like English now :P Stiil, wondering if there's any theory behind how fast these types of checks should be run. One thing is obvious, SMART is meant to expose a problem before it becomes a real problem of sorts. So, perhaps running the tests somewhat often is a good idea. But how often that IS the question. > > I have a follow up question for this. I can drop the /dev/sdX and use DEVICESCAN right? Absolutely > How would I tell smartd to stagger the tests if I have a 4+ drive raid? You mean slow down how fast a test is being performed? Ivan |
From: Christian F. <Chr...@t-...> - 2013-12-15 14:25:38
|
Ivan Lezhnjov Jr. wrote: > On Dec 15, 2013, at 4:58 AM, Weedy <...> wrote: > >> A very long time ago I found this and use it everywhere now. I don't even remember what the options mean anymore. >> /dev/sda -d ata -o off -S on -a -s (O/../../(7|3)/00|S/../.././12|L/../../6/03|C/../../7/03) -m root > I read all man pages, was a bit painful haha but I can read that line like English now :P You are welcome to provide (well tested and reviewed) patches(!) for the man page files. > Stiil, wondering if there's any theory behind how fast these types of checks should be run. One thing is obvious, SMART is meant to expose a problem before it becomes a real problem of sorts. So, perhaps running the tests somewhat often is a good idea. But how often that IS the question. It depends and may require some experiments. For servers running 7x24, one short test per day and one long test per week is a good starting point. For desktop systems running 5x8, long tests may not work at all because: - OS accesses may slow down the self-test such that it cannot be completed during working hours, and/or - the self-test may slow down the OS accesses such that the system is unusable. This depends on the vendor specific self-test implementation in drive firmware. >> I have a follow up question for this. I can drop the /dev/sdX and use DEVICESCAN right? > Absolutely Note that DEVICESCAN does not work for drives behind most RAID controllers. Possible implementations would be vendor and platform specific. >> How would I tell smartd to stagger the tests if I have a 4+ drive raid? Automatic staggering of tests is not possible yet. You need to add an individual smartd.conf entry for each drive. I agree that this feature would make sense. Please create a ticket: https://sourceforge.net/apps/trac/smartmontools/newticket Thanks, Christian |
From: Weedy <wee...@gm...> - 2013-12-16 01:44:21
|
On 15 Dec 2013 09:25, "Christian Franke" <Chr...@t-...> wrote: > > Ivan Lezhnjov Jr. wrote: > >> On Dec 15, 2013, at 4:58 AM, Weedy <...> wrote: >> >>> A very long time ago I found this and use it everywhere now. I don't even remember what the options mean anymore. >>> /dev/sda -d ata -o off -S on -a -s (O/../../(7|3)/00|S/../.././12|L/../../6/03|C/../../7/03) -m root >> >> I read all man pages, was a bit painful haha but I can read that line like English now :P > > > You are welcome to provide (well tested and reviewed) patches(!) for the man page files. > > > >> Stiil, wondering if there's any theory behind how fast these types of checks should be run. One thing is obvious, SMART is meant to expose a problem before it becomes a real problem of sorts. So, perhaps running the tests somewhat often is a good idea. But how often that IS the question. > > > It depends and may require some experiments. > > For servers running 7x24, one short test per day and one long test per week is a good starting point. > > For desktop systems running 5x8, long tests may not work at all because: > - OS accesses may slow down the self-test such that it cannot be completed during working hours, > and/or > - the self-test may slow down the OS accesses such that the system is unusable. > > This depends on the vendor specific self-test implementation in drive firmware. > > > >>> I have a follow up question for this. I can drop the /dev/sdX and use DEVICESCAN right? >> >> Absolutely > > > Note that DEVICESCAN does not work for drives behind most RAID controllers. Possible implementations would be vendor and platform specific. Mdadm raid, soon to be brtfs raid. This isn't a problem >>> How would I tell smartd to stagger the tests if I have a 4+ drive raid? > > > Automatic staggering of tests is not possible yet. You need to add an individual smartd.conf entry for each drive. > > I agree that this feature would make sense. Please create a ticket: > https://sourceforge.net/apps/trac/smartmontools/newticket Sure > > Thanks, > Christian > |
From: Nathan S. T. <nat...@on...> - 2013-12-15 23:44:56
|
On Sat, Dec 07, 2013 at 14:46:50 +0200, Ivan Lezhnjov Jr. wrote: > Per man pages long test can be run weekly, short test daily, automatic > offline immediate could be run per vendor default choice of typically > each 4 hours and so on. > > Well, how often should one run these tests? Any known best practices? > Recommendations? I'm not sure which (version of which) man page you found this info on, but when I originally started using smartmontools years ago I set up that same daily-and-weekly cycle, and I think I did so based on these two smartmontools-support list threads: http://sourceforge.net/p/smartmontools/mailman/message/7331341/ http://article.gmane.org/gmane.linux.utilities.smartmontools/1443 Subject: Recommended test and frequency http://sourceforge.net/p/smartmontools/mailman/message/7331590/ http://article.gmane.org/gmane.linux.utilities.smartmontools/1627 Subject: Testing Interval In short, there's no absolute "correct" answer, but this daily/weekly schedule is a compromise between catching failing drives fairly quickly and causing undue stress on the drives, and is seems to be appropriate for most environments. I haven't been following the issue particularly closely since then, but I haven't had any reason to change away from that schedule on my servers.... Nathan ---------------------------------------------------------------------------- Nathan Stratton Treadway - nat...@on... - Mid-Atlantic region Ray Ontko & Co. - Software consulting services - http://www.ontko.com/ GPG Key: http://www.ontko.com/~nathanst/gpg_key.txt ID: 1023D/ECFB6239 Key fingerprint = 6AD8 485E 20B9 5C71 231C 0C32 15F3 ADCD ECFB 6239 |
From: Ivan L. Jr. <iva...@gm...> - 2013-12-20 08:04:11
|
On Dec 16, 2013, at 1:44 AM, Nathan Stratton Treadway <nat...@on...> wrote: > On Sat, Dec 07, 2013 at 14:46:50 +0200, Ivan Lezhnjov Jr. wrote: >> Per man pages long test can be run weekly, short test daily, automatic >> offline immediate could be run per vendor default choice of typically >> each 4 hours and so on. >> >> Well, how often should one run these tests? Any known best practices? >> Recommendations? > > I'm not sure which (version of which) man page you found this info on, > but when I originally started using smartmontools years ago I set up > that same daily-and-weekly cycle, and I think I did so based on these > two smartmontools-support list threads: man page version: smartmontools-5.42 2011-10-20 SMARTD.CONF(5) I don't mean to say that there are actual recommendations in the man page, but there are examples that in my view were attempted to be sensible. > > http://sourceforge.net/p/smartmontools/mailman/message/7331341/ > http://article.gmane.org/gmane.linux.utilities.smartmontools/1443 > Subject: Recommended test and frequency > > > http://sourceforge.net/p/smartmontools/mailman/message/7331590/ > http://article.gmane.org/gmane.linux.utilities.smartmontools/1627 > Subject: Testing Interval > > > In short, there's no absolute "correct" answer, but this daily/weekly > schedule is a compromise between catching failing drives fairly quickly > and causing undue stress on the drives, and is seems to be appropriate > for most environments. > > I haven't been following the issue particularly closely since then, but > I haven't had any reason to change away from that schedule on my > servers.... Very interesting. Thank you for linking me to these. However, this makes me wonder now if these tests put a lot of stress on mechanics of the drives, why do vendors typically go for every 4 hours short self-test as is stated in the man page? Ivan |