hello world --

 

currently running 6.1.7 on my new server (yeah, i should post to "pimp your rig") with the following config:

* cpu: amd a8 7600

* mobo: asus a88x-plus

* ram: 8gb of ddr3 1600 ram (g.skill ripjaws, 2x4gb)

* hba: 2x supermicro aoc-sas2lp-mv8

* second nic: tplink TG-3468

* case: norco 4224, with 120mm fan backplate, all stock fans replaced with noctua ones

* usb thumbdrive:

sandisk cruzer fit, 16gb (d*mn, that thing is ridiculously small!)

* hard drives: all wd greens 3tb, except for two 4tb drives (one currently unused, waiting for v6.2) used for parity

*and* a 1tb wd red used for photo archive (a 3tb

* psu: 850w 80+ bronze evga supernova semi-modular

* ups: apc 1300va (BR1300G)

 

up until this past friday, i was quite happy indeed with this newly-built box: quieter, much more elbow room,

*faster* than the previous box i am in the process of lifecycling (will replace current innards with a config

similar to the new box when i have amassed enough money to do so), etc.

 

friday, when i came home from work, i noticed there had been a power outage a few hours earlier.  i wasn't worried

much about my server, as the current version of unraid has what appears to be a fully working apc ups daemon that

should really know when there's a power outage and to properly shut down the server if the outage is more than a

simple blip on the radar.

 

turned on the server, only to see one of the supermicro card was in a funk (could not detect the spin group/all

of the drives) and the server asking for a boot media.  whoa.  ok, reboot... kind of the same thing again. whoa, whoa!

 

long story short: the supermicro controllers ended up falling back on their feet after completely powering down

(unplug server from ups, etc.), but the cruzer fit had issues being bootable again, until this morning.  did test

the server hardware with bootable linux distros (mint, knoppix), looked ok, was able once to boot the flashdrive

to run memtest (no issues), but eventually had to reformat the cruzer fit and re-install unraid on it. 

 

the thumbdrive was still accessible from another windows box, so i just had to make a "stupid" backup (windows

explorer copy of the cruzer fit to a folder on desktop) and copy back the pro key once i had reinstalled unraid

and made the 'fit bootable.

 

even though i seem to have recovered from that power outage (i feel i should not have had to get agitated over

this), i now have three questions:

 

(1) anyone else had a similar issue with their flash drive after an apparent power outage?

 

(2) is there a way to see if the server did indeed shut itself down cleanly?

 

(3) after i assigned all the drives anew to their respective positions (s/n = drive 'x') and started the array,

the shares appeared by themselves, without any need for me to define them.  is a copy of the share information /

general configuration stored on the hard drives? 

 

thanks in advance.

 

hello world --

 

currently running 6.1.7 on my new server (yeah, i should post to "pimp your rig") with the following config:

.

.

.

 

* psu: 850w 80+ bronze evga supernova semi-modular

* ups: apc 1300va (BR1300G)

.

.

.

 

the thumbdrive was still accessible from another windows box, so i just had to make a "stupid" backup (windows

explorer copy of the cruzer fit to a folder on desktop) and copy back the pro key once i had reinstalled unraid

and made the 'fit bootable.

 

even though i seem to have recovered from that power outage (i feel i should not have had to get agitated over

this), i now have three questions:

.

.

.

 

thanks in advance.

 

It does not surprise me that the Flash Drive (contents only) was hosed.  That type of behavior often results when a write operation is interrupted by a power cut. 

 

The question is why was there a power cut?  One possible reason is shown in the edited list of your server's components.  I suspect that the total power required by the server exceeded the limit of what the UPS could deliver!  When this happens , the manufacturer has included circuitry in the UPS to instantaneously (or a half dozen milliseconds) shut down the UPS.  Remember when shutdown starts, ALL of the drives are spun-up at the same time in order to stop the array.  This means the power supply has to supply all of the starting current required by ALL of the drives.  You PS supply can supply a max of 850W but I doubt if that UPS is equal to the task. 

 

 

Quick look at spec for an APC 1300VA UPS showed that it is will only supply full load for less than 5 minutes.  What time are you using before you initiate shutdown?  Plus, as the batteries get older that spec time will be much, much shorter. 

the load on the ups actually appears to be surprisingly low: it ranges from 22% (171w) to 32% (249w), equating to a range of 24.1 minutes to 15.4, if the dashboard is to be believed. average load during a parity check (can't think of a better "acid test" for battery load) appears to be stabilizing at 32%.

 

as proof:

 

 

and

 

 

in theory, the system should have enough time to perform an orderly shutdown.  maybe i did configure the ups settings a bit too low, not giving enough time for a proper shutdown to occur.  i have now set it to 50% battery charge left and/or 10 minutes uptime left as triggers for a shutdown.  i am assuming that either value that is reached first will act as a trigger.

 

just in case someone might notice a stupid mistake i might have made:

 

 

 

i now have noticed something else in regards to my boot issues: despite the fact that i never had a boot issue prior to the outage last friday, now the bios insists on a eufi boot, which seem to be resulting in the cruzer fit not being recognized.  even after me reformatting that usb thumbdrive. when i select a "non eufi" cruzer fit in the bios boot menu, the server will then bring up the unraid startup menu.

 

i am now wondering what happened for the cruzer to no longer being bootable without me having to intervene.  either it's something in the thumbdrive, either it's with the bios.  quite puzzled by this, frankly.

 

Do not be confused by the running power when you have all of the disks spun-up.  There is a much, much larger current (and, hence, power) requirement when the half second or so when twenty-one disks spin-up.  I have seen estimated for start-up currents of much as 2 amperes (that is 24 watts) for some disks.  I do know that the green drives are much lower but many times exact figures are not supplied by the manufacturers.  As I said, there is a protection circuit inside of that UPS which will guarantee that it shuts down virtually instantaneously if the power requirement exceeds the set point.

 

One way to test is to simply pull the AC plug and see if it shuts down properly while watching a monitor attached to the server.  I believe you can do the same thing if you have logged on via a PUTTY session but I have never really check that out to verify it.  (Be sure that all  of the drives are spun down before you do pull the plug to get a test of the worst case condition.) 

@frank:

 

unless i totally misunderstand your post, from what i read, the ups currently used with my server should not let it cold boot at all because of the spin-up of all 23 drives requiring so much current for a second (or two).  or is there something i missed?  is it possible the supermicro controllers somehow don't let the drives spin up at the same time during the server's boot process?

 

@frank:

 

unless i totally misunderstand your post, from what i read, the ups currently used with my server should not let it cold boot at all because of the spin-up of all 23 drives requiring so much current for a second (or two).  or is there something i missed?  is it possible the supermicro controllers somehow don't let the drives spin up at the same time during the server's boot process?

i have 2 x SASLP-MV8 controllers and I know that during Post each controller seems to spin up its drives independently.  Not sure when the motherboard connected drives happen.  My UPS LED display shows that the current draw when spinning up all drives from the GUI after they have all been spun down is higher than the draw during the initial boot.

@frank:

 

unless i totally misunderstand your post, from what i read, the ups currently used with my server should not let it cold boot at all because of the spin-up of all 23 drives requiring so much current for a second (or two).  or is there something i missed?  is it possible the supermicro controllers somehow don't let the drives spin up at the same time during the server's boot process?

 

When your server is booting, it is running on the AC mains---NOT the battery and its inverting circuitry.  In this state, the maximum power which can be delivered to the server is determined by the protection on your AC supply lines.  When the power fails, it switches within a couple of milliseconds to deliver power from the battery and the UPS's inverter circuitry will generate the 110VAC which is then supplied to the output plugs.  At that point, the instantaneous power output (both VA and wattage) must always be less than what the ratings of the UPS are.  There is protective circuitry which will shutdown the inverter instantly to protect the switching transistors. 

 

EDIT:  Have you checked to see what happens if you pull the plug and let the UPS attempt to check the server down.  You have to actually monitor the shutdown process to see if the UPS is shutting down due to overload.  If that is not the case, you should be looking to see if the array can be stopped or is something keeping the array from stopping and the UPS is running out of battery.

(3) after i assigned all the drives anew to their respective positions (s/n = drive 'x') and started the array,
the shares appeared by themselves, without any need for me to define them.  is a copy of the share information / 
general configuration stored on the hard drives? 

 

shares are automatically created with default settings from the top level folders of the disks. Thats why your shares appeared by themselves.

 

  Remember when shutdown starts, ALL of the drives are spun-up at the same time in order to stop the array.

 

@frank1940  thats very interesting I didnt know that the drives spin up on shutdown.  So I guess even if the ups can handle that wattage if you set the ups to shutdown too late that extra power consumption could flatten the batteries before the shutdown is complete? So it would be best i guess to set ups to shutdown at say 80% battery level??

 

@tmp31416  I would start unraid and press the spinup all drives button and see what draw this puts on your ups

 

 

 

@frank1940  thats very interesting I didnt know that the drives spin up on shutdown.  So I guess even if the ups can handle that wattage if you set the ups to shutdown too late that extra power consumption could flatten the batteries before the shutdown is complete? So it would be best i guess to set ups to shutdown at say 80% battery level??

 

 

As a result of my experience, I have set my UPS's to begin shutdown 30 seconds after a power failure.  My observations have confirmed (that in my part of the world) that if it is out that long, it is not coming back on for at least a couple of hours.  This also provides for the shortening of the actual/true runtime as the battery ages.  It is best to remember that the 'Runtime Left' time is, at best, a SWAG.  A couple of days ago, I was looking at the specs for some APC UPS's in the same model series and observed that the full load runtime for a 1300VA unit  was about 5 minutes while the full load runtime for a 1500VA unit was approximately 4 minutes.  Why is the runtime less for the 1500VA unit?  It uses the SAME battery as the 1300VA unit!!!

@frank: i was actually thinking of changing my ups settings like you did (initiate a shutdown after 30 seconds on battery) as the computer might be wrong in its estimations of time left or battery levels.  encouraging to see i had the right idea.

 

to ensure the new settings will work, i presume i can type in zero in both the "battery level" and "runtime left" fields, only to type in "30" in the "time on battery before shutdown" field?

 

i now wonder if i should not lay my hands on a br1500g ups (the biggest unit i could easily get), as it is rated for 865w and is actually *expandable* (via unit BR24BPG).  unless i can find one of those rack-mounted ups like we have at work, for a reasonable price (yeah, fat chance there).  staying with the br1300g might not be a good idea, as the troubles i am now experiencing after this power outage are unreal. 

 

@gridrunner: i actually tried this, and the dashboard did not show more than 32% load on the ups.  or do you mean i should have the ups itself show me the load on it?

 

 

speaking of troubles, now i can no longer see my shares in the "shares" tab of the unraid menu.  they do list in the dashboard' shares list, they can be mounted from a client computer, but they do not appear in the tab one uses to define them.  what the...?  this situation has persisted over server reboots.  can't find any posting about this, am i the first one to experience this?

 

(and how do i fix this?)

 

btw, @gridrunner: i was talking about the previously-defined shares, the ones i had before reformatting the cruzer fit / reinstalling unraid & its key on it.  i thought their definitions re-appeared because they might have been stored on the drives themselves, but it was only the names (i.e. "hd_films", "sd_videos", etc.) that came back.  still unsure how.  furthermore, it turns out the drive assignments (share "x" uses drives "a, b & c", share "y" uses drive "d", etc.) didn't come back with the share names.  i only realized afterwards all share names were using "all drives" -- d'oh! i thus had to re-assign drives to their respective shares.  nasty surprise that was.

 

*edit* the shares configuration files do appear in the "config/shares" directory on the flash drive and, from a cursory look, do not appear to be corrupted.  so why the share definitions are not appearing in the web interface is quite a mystery to me.

 

 

so why the share definitions are not appearing in the web interface is quite a mystery to me.

adblocker? Try a totally different browser on a different machine. Try loading the webgui on your phone.

@jonathanm: the rest of the web interface works just fine.  the issue is in both firefox and chrome (which doesn't have script blockers.  firefox does, but i whitelisted my unraid server there). haven't tried with msie, though.

 

 

 

...

 

to ensure the new settings will work, i presume i can type in zero in both the "battery level" and "runtime left" fields, only to type in "30" in the "time on battery before shutdown" field?

 

...

 

 

I actually use 10% for Battery level

    and

10 Minutes for Time left.

 

I don't think it makes a lot of difference.  I hope that the 30 seconds will occur much earlier than either of those two settings.  If that does happen, the battery will be in very, very sad shape!

 

When you are looking at UPS's, you have to also look at the battery pack that is used in the ones you are comparing.  APC tends to use the same battery for a number of different VA rated units in the same family.  That means you get a bit more power out of the higher rated unit BUT it runs for less time at full load.  (Or the same amount of time as the lower rated unit if the loads are identical.)

 

Of course, when you are looking, you can often find a unit with a higher VA rating in the same family for less money.  Amazon is famous for this trick as they will use a specific model as a sale item at a bargain price for a short period of time. 

looking further in my boot issues, it appears the issue was not as much with the cruzer fit (i think), but with the motherboard's bios.

somehow, this apparent improper shutdown did have some nasty effects on bios settings, especially on boot parameters.

 

though the supermicro controllers were (briefly) affected, too. it's not just the motherboard.

 

so, on top of changing the ups settings in the unraid web interface, i will look for a better ups.  i have seen some refurbished apc SUA1500RM2U units being sold at an interesting price, will figure out if they are usable in a residential context (the reported 45+db noise rating is cause for concern)... and if i can run this unexpected expense past the girlfriend.

 

as for my shares not showing up in the web interface, i will look into this a bit further and start a different thread about it if i cannot solve this.

 

thanks all for the feedback.  cheers.

 

*edit* trying with msie (the only browser i hadn't tried yet), i now see the shares in the web interface.  i will have to check if extensions in my other browsers (firefox & chrome) aren't somehow blocking my unraid server, despite it being whitelisted.

 

*edit, part 2* one of my supermicro cards (it was never both, just one) can still be giving some episodical errors on boot, that are cleared by a reboot (for now).  that episode from last friday has indeed been having some weird effects on my server.