Guzzi

Hi @bonienl, I think I have read it before somewhere, that on the dashboard, the CPU frequency is NOT the current speed - while e.g. fan speed IS current. Is it possible to change that, so the user can verify, if powermanagement works properly? I was successful to get the current CPU frequency with cat /sys/devices/system/cpu/cpu*/cpufreq/scaling_cur_freq What do you think?

The plugin scans the system in order to populate the PWM controller list automatically. If it stays empty then nothing could be found. It uses the command: find /sys/devices -type f -iname 'pwm[0-9]' You may want to try manually to see the results. Hi bonienl, it seems, I have some trouble with the autodetected PWM controllers. when I check the system with pwmconfig, it tells me to use /sys/class/hwmon/hwmon1/device/pwm2 for fancontrol (which worked for me in unraid5 and fancontrol script in go script). The dropdown to PWM2 in autofan however uses the PWM2 selection from device tree, resulting in following command line: /bin/bash /usr/local/sbin/autofan -c /sys/devices/platform/it87.552/pwm2 -f /sys/devices/platform/it87.552/fan2_input -l 75 -t 31 -T 41 -m 3 The reading of the fanspeed works fine - but the setting does not. Any recommendations?

Hi DoeBoye, sorry for the trouble you had - 1015 controller was the exact reason, WHY I removed that ata option from the script for myself ...

+1 .... currently always manually deactivating cachedirs before starting parity check ;-)

Hi JoeL., what do you think about an output to the log file, if cachedirs takes longer than a certain amout of time repeatedly (let's say more than twice - this filters the first time reading and any rereading when the array was accessed and dcahes flushed) ? I think that would help a lot to identify the case, when it is not working properly and "overloaded" - instead of trying to observe that by checking spindown behaviour. If it is not taking too much cpu cycles, such an output could include the number of files read by find - so user get's an info with what number of files that case happens - what do you think? Doable?

Yeah, I understood that - I was asking for a real number (maybe during your experiments) with stock settings - as I could then do a count (assuming the size per entry is known) on the data and get an idea how much is possilbe at the maximum - and then exclude/include only some areas for cachedirs script ...

So is there a known limitation in terms of "files/directories" that can be cached with current 32bit architecture?

good hint ... google gave some more info, e.g. available memory (not only dentry) grep -i slab /proc/meminfo memory details: awk '/dentry|inode/ { print $1,$2,$3,$4}' /proc/slabinfo live memor usage slabtop -s c JoeL, Weebotech - would it help to check dentry usage there, while cachedirs fills buffers/dentries? (I hardly understand what all this means and where/what to check .....)

Thanks JoeL. for the quick answer - so back to "try and error" ;-) - I think, I have too many files there in some areas, that just flood the buffers and flush old ones (I noticed a LOT of reads on stat page after some time with cachedirs active recently - and only a few without, so I think that is a clear indicator...). Is there a rough formula how much (buffer) memory is used per directory and per file/dir entry - that would allow to check, if there is a general "sizing" problem?

yeah, those script versions are spread around the forum - and hard to find ;-). As aiden said, script has evolved over time by several contributors - and the one I posted above was "the best" to my knowledge - mostly, because I a) do not need to change something, if drives are added or removed from array and b) for the continuous rpm regulation (that you also implemented). iirc, JoeL just recently improved unmenu to adaptively use -D ata parameters when checking drives - that might be worth to include, if it can be automated without causing regression - didn't check what he did for that - but would make the script more universal for to old hw and e.g. newer sas controllers (usage with or without "-D ata" param)

Hi JoeL., is there any command parameter/setting, that would make sure, writing to the array does not clear/loose the cachedirs buffers? Currently, after I copied something to the array, cachedirs has to reread a lot from disk again. And one more question: Is there a way to see, how much memory cachedirs (ot "the populates cache buffers") occupy - to see, if there is not enough memory?

Hi Pauven, thanks for your contribution. Just as a note: There is another fan control script "somewhere" on the forum (can't remember where, but I am using it since a long time), that does automatically detect the number of drives in array and also sets the pwm speed in steps between off,min,max plus support for command lines and can be run from go script, not requiring a cron entry) - I attach it here if you're interested to look into it - all credits go to the authors fancontrol.sh.txt

My server takes approx 2,5 Watts when in S3, so yes, RAM refresh is required but doesn't need very much. If you loose power, you cannot resume as you stated, it's like disconnecting while the system is running. you should use a UPS anyway ... Might be worth to call a "sync" before going to S3 - I have never done this so far.

Hmmm, I remember theat I have read this post in the past - but didn't try because of the warnings. Currently have no testsystem to give it a try on not in use environment ... Googling showed up posts that people were not able to set back drive to original state (WD-drives, as I do have them). So before giving this a try I would be interested in experiences of other user and better understanding of what it means. - hdparm -s sets the POIS mode of the drive - On cold startup this causes a very long delay (approx 15 sec per drive = 300 secs for 20 drives = 5 mins !? - It seems, 3132 is supporting this - no info found if also drives via portmultiplier word (i assume yes) ; found no infos about drives connected to SB700 southbridge as well - On WOL (wake-up from S3 mode) the consequence would be, that the drive would NOT spin up (and immediately fall asleep), but just switched online controllerwise without spinning - right? - If the drive is accessed by UnRaid, the drive will spinup as usual Can anyone confirm or correct me - and also feedback to user experiences are welcome.

How to you do that? My WD-Greens spin up when server wakes up via WOL and after "reinitilization" - some seconds later - spin down again. Can you point me to some info regarding those jumpers to avoid spinup after WOL? I thought this is required for reinitialization of the drive. And I also have several resets after WOL to reinitialize the interfaces, e.g.: Feb 11 22:18:28 XMS-GMI-01 kernel: ata3.00: SATA link up 3.0 Gbps (SStatus 123 SControl 320) Feb 11 22:18:28 XMS-GMI-01 kernel: ata3.00: failed to IDENTIFY (I/O error, err_mask=0x11) Feb 11 22:18:28 XMS-GMI-01 kernel: ata3.00: revalidation failed (errno=-5) Feb 11 22:18:28 XMS-GMI-01 kernel: ata3.15: hard resetting link Feb 11 22:18:28 XMS-GMI-01 kernel: ata3: controller in dubious state, performing PORT_RST Thnaks, Guzzi

A lot will depend on how many files & folders you have... basically, you are now doing a "find" on twice as many. It might even be worse, as the entries from the disk shares you want to stay in the cache, to keep them from spinning down, might be displaced by the extra ones read by scanning the user share's twin copies. so this is the result after running the box overnight: All disks are spun down and memoryusage is as follows: root@TOWER:/# free total used free shared buffers cached Mem: 4017452 3866908 150544 0 247984 2939772 -/+ buffers/cache: 679152 3338300 Swap: 0 0 0 I ususally access the server for files via the [TOWER-Root] - path = /mnt/user entry in smb-extra.conf to get all usershares on one drive; this access (from windows SMB1) is fast and no disks spin up. Can you comment on how to read the output of free-command? I check the same without -u parameter to compare later... THanks, Guzzi

Thanks for explanation - I currently have it installed and checked with htop - so CPU usage seems pretty high, but can't say if it was the same before without -u. Will just let it run some days and see "how it feels". rgds, Guzzi

Hi Joe L., I have another question: What is the usecase for the "-u" parameter? I understand it reads and caches the usershare - but I thought, this is already working, if the disk dirs are cached? If -u is enabled - would the excludes still work? (Because I have mapped dirs in the usershare, that I do not want to cache) Thanks, Guzzi

THe only way is to test several improvements like BIOS settings, HW, etc. to find a working config. Might be possible, that your hardware does not support S3 properly and thus there is no working solution/workaround - then it would be the only solution to change HW - that's the way I did in the past ...

When I upgraded the bios I did a "load optimized settings" I did not overvolt my CPU, was not necessary - but I removed the undervolting because I experienced instabilities similar to yours. Overvolting (minimal!!!) nevertheless might help to stabilize the rig in some cases - that's why I mentioned it. So loading the default usually should be fine - you might want to load "optimized", although this is usually using "tighter" settings, it might better for the rig - unfortunately there is lots of bad BIOS implementations everywhere :-( - I assume this is the reason why Tom officially is not willing to support S3. As a solution it would help to add a testsection for S3 and point working setups - I would have been very happy, if this would have been available - avoiding all those "try and errors" by myself ... btw - PSU should be fine. I have no experience with your adaptercard. There is always hope to get implovement with newer kernel - the last update made quite some improvements to powermanagement. To make it even more complicated: Even n older BIOS might work better ... I know this sounds bad, cause means lot's of testing, but unfortunately I have no better hints at the moment ...

not easy to answer this ... I have NOT have this experienced ever (so far) in my environment, using S3 regulary. I assume, you have no entries in the syslog, so the freeze is before any reinitialization happened!? Possible reasons (fixes) could be: - Powersupply too weak (voltages drop when WOL happens and MB/CPU hangs) - check with no drives attached or check with "sifficient" powersupply - MB HW incompatibility - change MB (I did so in the beginning, 'cause I wastn't able to get S3 properly running with my former MB) - ACPI/BIOS incompatibility - check if newer/fixed BIOS is available or try using addons circumstancing that (you already did some stuff like boot parameters, e.g. I have to use S2RAM to properly reinitialize graphics after WOL. There is lot's of stuff in the linux kernel to handle bad BIOS/MB/ACPI behavior, so also a newer linux kernel might fix it/help. - instability by configuration: If you are undervolting, this could lead to freezes in such situations - use standard config or even try a little (!) overvolting to stabilize the set (I used undervolting in the past to save cureent, running on standard now plus S3 as better solution and avoiding freezes or other problems. You could also try resetting to defualts in BIOS to get other settings back to normal and see if it helps. Difficult to solves such issues - as stated above, I had quite some problems in the past and ended up changing the mobo - my final solution was to order some mobos and test them - I kept the one working best and sent back the others. Because I am a linux noob, other people can help you more on linux specific things happy new year and "happy testing", Guzzi

... maybe your drives do not spin down because they're not in the array? You might want to check this - as far as I remember the script only sends the server to S3 if all disks are spun down.

I'm running the same versions. I only cache three levels deep. I only have 512 meg of RAM in my server and under 5000 files in my Movies and Music folders. You have lots more files you are attempting to cache, but far more memory too. Yes, it is normal to see two processes. part of the script is run in a sub-shell. (The while loop is run as a sub-shell) Hi Joe, I also have lots of files and directories on my server and experience a lot of reading on the disks. I upgraded memory to make sure, it is sufficient to hold all dir-infos. Nevertheless my disks DO spindown (all), but often reread all contents. Might be caused, when I write to the array so the buffers are reused for writing? So my question (I am Linux noob....): Is there a way/tools to check/monitor what's happening? currently I used "free" to check memory usage and the mgmt page to see read operations on the drives. So I can see how many reads I had for first run of cache_dirs - and after some days of operation I see 10 or 20 times the amount, although there was no "real" read on the disks. Thanks, Guzzi Edit: I use S3 and wake the server via WOL - does cache_dirs a reread after S3? Does UnRaid flush the buffers after S3-wake? Hmmm, that would explain it - how can I verify this and is there a way to avoid it? Here is my "free" output: total used free shared buffers cached Mem: 4017452 2313480 1703972 0 248792 1391780 -/+ buffers/cache: 672908 3344544 Swap: 0 0 0 Do I read right, that mem extension from 2 to 4 GB wastn't necessary?

Anyone have a clue why this may be? Most likely it's due to a bug/feature in the MB BIOS implementation of S3. Seems to be. After flipping ACPI bios settings willy nilly, it is now working....mostly. It doesn't re-post video on resume, but I don't care. It's a headless system now that its running anyhow. I had the same problem and solved it by using S2RAM.

100% agree with agw! forget about all the scripts and check, if your machine supports S3 properly (by far not all HW/BIOS combinations do!). Entering S3 from commandline: echo 3 > /proc/acpi/sleep pressing powerbutton should then wake up your machine - check syslog if everything is fine. If yes, at least S3 mode entering worked fine and reinitializing too. Second check, if waking up the server via magic packet (instead of power button) works by sending a magic packet to your NICs MAC address if all this is fine, you can try integrate those scripts - otherwise you won't be able to use them anyway with your setup. and btw: S3 is not officially supported by Tom/limetec - probably because of the problems with so many HW/BIOS environments I assume - I had to change my MoBo to get it working ;-)