And, its the same time suck....cacti.
Last weekend got away from me, because I to make another attempt to improve
cacti performance. I had tried adding 3 more devices to it, and that sent it over the limit.
I tried the
boost plugin....but it didn't help, and only made things more complicated and failure prone. Evidently, updating rrd files is not a constraint on my
cacti server. Probably because of running on an SSD.
I made another stab at getting the percona monitoring scripts to actually work under script server, but that failed. I suspect the scripts aren't reentrant, because of their use of global variables and relying on 'exit' to cleanup things it allocates or opens.
I had blown some previous weekend when I had tried to build the most recent version of hiphop to maybe compile the scripts, but after all the work in figuring out how to compile the latest 2.0x version...it would SEGV, just as the older lang/hiphop-php did after resolving the problem of building with the current boost (a template had changed to need a static method, meaning old code won't link with newer boost libraries without a definition of this.) And, this is beyond what I have in my wheelhouse to try to fix.
During the week, I had come across some more articles on tuning FreeBSD, namely a discussion of kern.hz for desktop vs servers. Where it being 1000 by default is good for desktops, but the historical setting of 100 being what to use for servers. Though IIRC, ubuntu uses 250 HZ for desktops and 100 HZ for servers, it also doesn't do preemption in its server kernel along with other changes (wonder if some of those would apply to FreeBSD?) Though modern kernels have been moving to be tickless. Which I thought was in for FreeBSD 9, though the more correct term is dynamic tick mode...and which is more about not doing unnecessary work when things are idle. Which isn't the case with 'cbox'. So, perhaps, fiddling with kern.hz and other sysctls might still be relevant. Though haven't really found anything detailed/complete on what would apply to my situation.
So, I thought I would give kern.hz=100 a shot.
At first it seemed to make a difference....no improvement in how long to complete a poll, but the load was lower. Until I realized that a service had failed to start after reboot. I had only run the rc script by hand, I hadn't tested it in a reboot situation. And, its not an rc script....it was used to be a single line in rc.local that worked on ubuntu and FreeBSD (except on one of the Ubuntu systems it results in a ton of zombie processes, so making it an init.d script that I could call restart on happened.
So, I spent quite a lot of time reworking it into what will hopefully be an accept rc script. One thing I had changed was that instead of using a pipe ('|') which was causing the process after the pipe to respawn and turn the previous process into a zombie each time the log file was rotated and "tail -F" announced the switch. And, this was while I was moving the service to FreeBSD (and management under cfengine 3.)
Though looking at my cacti graphs later....while the service had failed to start after reboot, it turned out to have been running for sometime, until I had broken it completely in trying to rc-ify the init script. Will, duh....I had cfengine set to promise that the process was running, and it had repaired that it hadn't started after the reboot.
Another thing I had done with I had init-ified the startup of this service, was I switched from using pipe ('|') to using a fifo, which addressed the respawning and zombie problem and eliminated the original reason to have an init.d script....
While the init.d script had worked on FreeBSD...it was just starting the two processes with '&' on the end then exiting. FreeBSD's rc subroutines do a bit more than that. So things weren't working. The problem was that even though I was using daemon instead of '&', so that daemon would capture the pid and make a pidfile. seems daemon wants the process it manages to be fully working before it'll detach. But, the process is blocked until there's a sink on the other end of the fifo. (does sink fit was the name for the fifo's reader?) I first wonder if I could just flip the two around, but I suspect starting the read process first would be just as blocked until the write process is started. So, I cheated by doing a prestart of the writing process and only tracking the reading process.
Though it took a bit more work to get the 'status' action to work....eventually found I needed to define 'interpreter' since the reading process is a perl script. And, the check_pidfile does more than just check to see if there's a process at the pid, but that its the right process. And, it distinguishes between arg0 and the rest.
Pretty slick...guess I need to do a more thorough reading of the various FreeBSD handbooks, etc. Of course, it has been 13+ years between when I first played with FreeBSD to its take over of my life now.
As for the tuning....it had made a small difference, but no improvement on cacti system stats. Basically the load average fluctuates a bit more and the CPU utilization seems to be a bit lower...though it could because the 4 lines of the cacti graph aren't so close to each other now.
Meanwhile...I noticed that one of the block rules in my firewall had a much higher count than I would expect, so I think I was about to get logging configured to see what that's about.....(which I was working on when I remembered that I hadn't rebooted after making the kern.hz change to /boot/loader.conf yesterday...the commit also picked up files that I had touched while working on moving the one remaining application on 'box', though that may get delayed to another weekend....perhaps the 4 day one coming up.)
I had set cf-execd's schedule to be really infrequent (3 times an hour), because I was doing a lot of testing and cf-agent collisions are messy....messier than they were in cfengine 2 (in 2 it usually just failed to connect and aborted, in 3 it would keep trying and splatter bits and pieces everywhere....which is bad when there are parts using single copy nirvana. resulting in services getting less specific configs, until the next run.
But, I sort of brought back dynamic bundle sequences.... but key off of "from_cfexecd", so I can test my new promise with less problems of colliding with established promises. Though there are other areas where things still get messy.... need to clean up some of the promises I had based on how things were done at work, so that the promises are more standalone.
Kind of weird using my home cfengine 3 setup, and other admin activities, as the means to break the bad habits I had picked up at work....
I had acquired my first HSTI Wireless Media Stick back on April 24th, 2011 (from a marketplace seller on Amazon.com)...it took some time to arrive and I blogged about it on April 30th, 2011 -- Getting local content to show on my Roku XDS.
Now my Roku has moved to my other HD display (24" 1080p), but that was before my old Samsung HDTV (43" 720p) regenerated into a Samsung Smart 3DTV (50" 1080p).... so I'm back to living room TV being my main viewing device for all content, though TiVo has a box that I could connect to the smaller display to access my TiVo content there....which I may want to get at a later date. The majority of content I watch is from TiVo...one of these days I need to setup my blu-ray player so I can get back into watching DVDs (not sure when I'll have blu-ray discs, but need to get back on my netflix backlog).
But, the other day I had an mp4 file that I needed to play....and I thought I should get someway to do that to my 50" HD display... Had to settle with using the Roku for a bit. And, decided that the plan will be to acquire another HSTi Wireless Media Stick.
After searching around online, eventually found that ordering directly from HSTi was the only option now. So, I ordered another one on May 17th. It arrived yesterday. But, I didn't set it up until I got home from work today. Somehow I had forgotten again that HSTi is in Calgary, Alberta. Not that I'll be going up there in the immediate future....
Anyways, no big surprises...good thing I had solved my USB2.0 and Windows 7 in VirtualBox on FreeBSD problem (got a Silex SX-DS-4000U2). I'm sharing TARDIS from orac to it still, since I don't yet have a replica on zen yet (need to free up space). Though when I moved the HSTi Wireless Media Stick it had forgotten the share, so had to pull up web again and add it back. Interesting that its graphic shows the itself, while the graphic on my older stick is that of the original Wireless Media Stick (it used to be the correct graphic, but after an update it keeps showing the graphic of the older version.) Though this one came from the factory with the latest firmware, so who knows what'll happen when there's an update.
Was interesting using the SmartTV to view it, though wonder if it'll be a problem with it constantly discovering the stick every time I turn it on and presenting dialogs and such. Afterwards I tried the Amazon app to see if that was working yet....it was still saying I needed to update my TV, though this time there was an update....and now that works. Which might make it interesting to decide on what I should do. The only problem with using the SmartTV versus some other viewer....the TV is only 2.1 audio while other routes I can get 5.1, and its a different input on my receiver....
Oh well, back to other projects....
I have a number of FreeBSD projects at home on the go that I've been trying to find enough free time each weekend to tackle. Having a recent problem where my system will freeze/panic/reboot whenever I shutdown or saved a VM...I thought maybe upgrading to 9.1 would help. It didn't.
Anyways...I wondered how freebsd-update would proceed, given that checksums on everything had changed...because at one point during things I had rebuilt the world with optimized compile flags, though otherwise still running GENERIC and no other alterations, though later rebuilding kernels so uname would reflect p4 and p5....
But, from what it said it was going to do to get me upgraded, it seemed okay with everything. Though the merge was a bit strange....first time through it didn't present one of the merge conflicts for me to resolve (/etc/login.conf), so I said no when it asked if the file was ok before proceeding. That just stopped it instantly, had hoped that since it didn't ask how to resolve before that it would ask now. But, running the update again....it did ask for that file this time (otherwise I was considering what would happen if I fixed it by hand during or later.... So, then it went and upgraded.
And, then that's where all the fun started. While rebuilding all my ports wasn't necessary. It was necessary to rebuild all the ports that create kernel modules....like nvidia, virtualbox, fuse, webcam....
Over the last year, I started buying SSD drives. It used to be that they seemed pretty expensive, and of questionable performance and reliability. But, all things have improved over the years. And, when 120GB drives dropped under $1/GB (initially after rebates, later before rebate [if any]). I didn't have an immediate need for an SSD drive at the time, but I envisioned replacing the drive in my (u) laptop. And, perhaps my (w) laptop....beyond that I wasn't sure.
3/05/12 - Patriot Pyro 120GB - $159.99-$40 rebate = 119.99 3/20/12 - OCZ Agility 3 120GB - $139.99-$30 rebate = 109.99 4/27/12 - Mushkin Enhanced Chronos 120GB - $ 99.99 5/16/12 - OCZ Solid 3 60GB - $ 74.99-$20 rebate = 54.99 8/10/12 - Sandisk 128GB - $ 79.99 11/21/12 - Kingston HyperX 2K 240GB - $149.99 1/18/13 - Sandisk Extreme 120GB - $ 89.99-$15 reward = 74.99 1/18/13 - Sandisk Extreme 120GB - $ 89.99-$15 reward = 74.99
But, during this time...I had the lhaven misstep Where I had picked up the 60GB drive for that, but ended up using the Mushkin 120GB drive instead. The OCZ Agility 3 120GB had gone in as OS drive for my Xen Cloud Experiment. And, stayed when I went on to making it FreeBSD. Cut up initially as 64k boot, 32GB swap, 16GB L2ARC for the mirrored 1.5TB drives to help with dedup...and the rest...63GB root zpool.
Somehow the Patriot got misplaced for a while, so it got overlooked during the chaos.
After a 'break'. I picked up the Sandisk 128GB drive. Thinking that it might be a better choice to use to finally replace the (u) laptop harddrive. Though I waited until after the NN conference in October to do it, but before I went to LISA in December. Though I didn't finally upgrade the the OS to 12.04 (from 10.04) until just last week. I had thought about doing a clean install to 64-bit....since there were some issues since I upgraded the memory to 8GB. But, changing the hibernate method seemed to have solved the issue....so decided to leave it 32-bit. Though my (w) laptop is 64-bit...though it only has 4GB. Not sure if I'll upgrade it to 8GB. Or when I'll upgrade its harddrive to SSD.
Things have been kind of tight since...on the 128GB Sandisk drive. partly because swap got a bit bigger...had suspected that 8GB swap was iffy for hibernation....so had bumped that up. Plus the original harddrive was 160GB. But, the lion share of space consumption is my Windows XP VM. But, it get's the job done.
Meanwhile...around this time I got the idea that instead of making the risky upgrade of my two Ubuntu servers from 10.04 to 12.04, that I would set up two new FreeBSD servers and migrate the essential services over before deciding the future of those systems. So, I acquired a pair of Shuttle XS36V's...4GB of memory for each...and then eventually the plan was to acquire a pair of SSDs for them. Which I finally did last week as a pair of SanDisk Extreme 120GB drives. (with the help of $30 from Best Buy Reward Zone....and this purchase should get me a $5 reward zone soon.) And, these will probably get installed as FreeBSD 9.1.
Pages: 1· 2
This was originally going to be a very long post, but I kept putting this off ... and now I just feel that something needs to be said.
The story starts with waking up on February 15th, to find zen was dead. It had self updated overnight, and now it was unbootable, and the start repair couldn't get me back. Apparently, the problem had started long ago with all the previous times where Windows 7 would lock up...usually under intense disk activity...and the afterwards, the intel matrix raid would require re-initialization of my 1.5TB RAID 1 array.
Apparently, it was slowly corrupting my drive....because trying to restore from WindowsImageBackup was also a failure. Since this happened the day before Gallifrey One, I had to wait until I got back to do some more serious attempts to recovery, during which I ordered a full copy of Windows 7 Professional, hoping that a repair install might be an option. It isn't because the repair option can only be invoked inside a running Windows 7 system .... in need of repair. Not by booting the disk. ARGH!
At least I should have the data in BackupPC to restore from.....though hopefully before the bit rot of its ext4 filesystem makes it go away. Plus I had hoped to get some configuration going where I could mount the RR62x RAID 5 array, and get at the Oops!Backup store.
So, the plan now was to wait until Ubuntu 12.04LTS to land and then maybe some configuration of running Windows 7 in VirtualBox and recovering into that, etc.
For a long time, I've been running a 6 drive RAID 10 array of Hitachi 5K3000 2TB drives in Orac for backuppc. This configuration got me at 5.4TB array, and somewhat better performance than when I tried a RAID6 configuration. But, eventually, I kept running out of space and the price of harddrives went up so expanding the array over time didn't happen as I had hoped. Being RAID10, the options were other concat another array, either 2 in RAID1 or 4 in RAID10 or 4 as 2 RAID1....using volume manager. Or maybe see if RAID10 would deal with having all 6 drives upgraded to 3TB, though hadn't considered the transition of 512 to 4k and how it would cope with that.
Though I did, eventually find out when I upgraded a 1.5TB RAID1 set to become a 2TB RAID1 set....going from ST31500341AS to ST2000DL003, where I contributed my experience here: http://askubuntu.com/questions/141669/creating-properly-aligned-partitions-on-a-replacement-disk/ Though it first started because one of the ST31500341AS drives had failed.
Before the failure of one of the 1.5TB drives in the above mentioned RAID1 set, I had 4 ST31500341AS in a RAID5 on old-Zen. It had been done in under RR622, under Windows, and NTFS partitions, etc. I had tried copying the data at various times, not really having anywhere that would hold the data elsewhere...but wanting to get it over to FreeBSD for recovery. While I got the rr622 driver working, and it saw that I had a single array (rather than the native driver that would see the 4 individual disks.) I couldn't get access to the data. Though it had worked when I was previously playing around with Xen (had tried copying it then....to a 2 1TB RAID0 set, but then one of the 1TB drives died....so I lost the copy, I had then replaced it with a 2TB RAID1 set....using an ST2000DL003 and an ST2000DM003, the DL being a 5900RPM drive and having a 5 year warranty...while the DM drive is a 7200RPM drive, but with only a 1 year warranty. And, turns out the 1 year is generous.
At work, I had built my FreeBSD desktop using a pair of the ST1000DM003 drives...and 3 drive failures later....it is now a pair of ST2000DL003 drives. Yeah...I was having trouble with the array, and apparently using XFS was a mistake too...because I thought it was recovering, but instead it was slowly eating the data. When I had nuked the RR622 RAID5 array, and had switched to using it as JBOD and create a RAIDZ set under FreeBSD...I found that there was nothing to copy back from the RAID1 array. D'Oh!
Though I had also copied the Microsoft WindowsImageBackup files, to see if I could mount the VHD file under VirtualBox to help in recovery. I largely had the data in bits and pieces elsewhere, it was the environment I was wanting to recreate...and Oops!Backup didn't back up that part anyways (the data I was mainly trying to migrate). The image mounted, and I could see it...but soon after Windows would try to fix it and then it would disappear....kind of like what it did on February 15th to make the original Zen go away. No idea what kind of disk rotting the Intel Matrix RAID had been doing, when it had to initialize the array again every time after a Windows crash. I've had Ubuntu crashes, but the RAID arrays remained stable...usually. While Windows & Intel RST....it was pretty much every time. I'm sure it was slowly corrupting things overtime to where things wouldn't recover, though it choose to do that after an automatic reboot for Windows updates...and the day before I left for my first Gallifrey One made things even more annoying.
Anyway with another 1.5TB drive freed up, I contemplated adding it to the RAIDZ I had made of the 4 1.5TB drives, keeping it as a hot spare, or just use it by itself -- living dangerously. I ended up with the latter for some temporary data. Because in my mind I was starting to lean to what happened next.
Pages: 1· 2
So, a while back I read that it would be possible to upgrade a number of older computers from spec'd maximum to a bit more.
Officially, the maximum supported memory of the Lenovo 3000 V200 (ulkc) is 4GB... My other laptop also supports a maximum of 4GB, but I suspect that between them, I should be able to upgrade one or both to 8GB. Though I prefer to have it work in this one, since I use this one more.
So, a while back I had a newegg coupon to get laptop memory, so I got a pair of the SO-DIMMs. But, then I set them aside....and forgot about them.
More recently, I picked up a new battery for this laptop...and had it in the bedroom to run down and see how much life I would get. Online they sell aftermarket batteries for 4800mAH and 5200mAH. Though it turns out the original battery was 5200mAH.... So, that's what I have....anyways it works. And, I let the laptop rundown. So, I was going to try the memory swap after that. Though that was a few days ago....
So, lhaven was this computer I bought specifically for Stanford's Quake Catcher Network project. This was because when I received my sensor, I found that it only worked on Windows or Mac. Linux would require the older sensor, which they weren't selling anymore.
Things seemed to work...except that the machine would lock up now and then. Well, it is Windows...so maybe I'll just schedule it to reboot regularly.
At first that seemed to help...but gradually, I was rebooting it more and more frequently. Eventually, I was down to rebooting it 5 times a week. Sometimes it would fail to restart, other times it would still lock up between reboots.
Eventually, they came out with Linux support for the sensor and I moved the sensor over to orac. That worked great for a while...and then it got a kernel update, and then starting VirtualBox would cause the machine to reboot. And, not just when I needed up update my Quicken. Because I was also participating in LHC's Test4Theory.
So, it was time to find a new home for the sensor. After moving it back to lhaven.
I looked for various low power x86 dedicated SBC or plug computer to run it. Though these things aren't cheap, so I'd want to do more than just QCN with it.
Then it occurred me...that I was overthinking this whole mess. Why don't I just install Linux on lhaven?
I haven't been looking forward to the leap from 10.04LTS to 12.04LTS. While Orac has a GT440 in it that's not being used, not sure what I would do for box....though I've been eye'ng either a GT218 or an NVS 300. It only has an LP PCIe x1 slot.... there are other concerns.
In the Scheme of things...I had planned to upgrade the video card in zen someday...and take its GT420 and change to an LP bracket and put it in lhaven. Hadn't quite worked out what I'll do about gumby, but it has since died...and when I get around to resurrecting it, I might have a Radeon HD5450 that I'll free up (not sure if it helps with BOINC, but eventually...I want all the systems using DVI....so I can switch to my new 4-port DVI KVM. DVI to VGA through the old KVM was horrible....plus keyboard issues with the old KVM, makes me wonder if I can wait until I can fully make the switch.)
So, I had downloaded 12.04LTS beta1, but never got around to install it anywhere....and when this idea occurred, 12.04LTS beta2 was just starting to appear. Though the question was, would I go with ubuntu, or try one of the other variants. Namely, xubuntu or lubuntu.
I ended up going with xubuntu 12.04LTS beta 2. I didn't hear about lubuntu until after I had started downloading xubuntu...so I envisioned that I would resurrect gumby with lubuntu....though so far I haven't gotten around to that.
I was just wishing, I had a little bit more space in my dropbox....when I came across this.
This is also gives me the excuse to play around with my new camera lens. An smc Pentax DA 35mm f/2.4 AL.
Wonder if I should still take both lenses with me when I go to Gally?
After which, I learned that a 58mm cap fits the end of the hood nicely....so I bought a Promaster SystemPRO White Balance Lens Cap - 58mm.
But, this then means that the Camera Buckle Lens Cap Holder; for 62, 49 and 40.5mm Lens Cap Sizes that I had gotten wasn't sufficient anymore. Though it never really was, given that one side is blocked with on my strap...so I guess it really worked out that I needed to add another one of Camera Buckle Lens Cap Holder; for 67, 58 and 52mm Lens Cap Sizes to my strap as well.
62mm is the cap size of my Tamron AF 18-250mm F/3.5-6.3 lens....which is now the other lens that I'm debating if I should bring on my trip....
Meanwhile...I'm wonder where one finds a lens chamois these days.... Think I at least have a recent lenspen somewhere. Also I discovered that the online storage part of my Eye-Fi cards aren't working....probably has to do with the great password reset at DreamHost.
http://db.tt/8sC6ZtO <-- Bonus .5GB signup link.
Winter has been weird this year...but there's a winter storm alert finally? Though its the rain and snow kind with ice accumulation and inches of snow type. Just as long as it isn't another icepocalyse of 4 years ago.
Though seems interesting that the computer I bought to replace the one I lost during the icepocalyse, died on Friday. Almost as if it was designed to die after 4 years. Annoyed, because its orac, my backuppc/file server. There's 16 drives on the machine, a total raw storage of 23TB....though only about 11.5 usable TB. And, a whole lot less in TiBs.
I'm hoping its something simple though....was looking at the hardware monitor in the BIOS screen...where it reads out voltages...and the lower voltages are low....though are those from the supply or are those generated on the motherboard? Seemed to me that motherboard generates some of the values, but not sure what I'll do if that's the case.
Also seems odd that it died shortly after I turned on heat Friday morning. Anyways...I had been considering getting a new power supply for orac for other reasons, so I jumped over to Amazon and took the plunge to get a new one by Amazon Prime. I had also been meaning to upgrade the case fan on it orac....I mentioned this before....stock fan is 80mm, but there's holes for a bigger fan. I had/have some extra ones around, so I had tried those...except that my extras are too big...so not 120mm. So, I had looked at 92mm, and I did end up with one, but failed to notice that it wasn't a 3-pin fan. So, maybe I'll add one of those to.
So, cart is adding up...and I'm looking at items I have 'saved for later'....and well, new humidifier is among there. Its pretty expensive....~$88 Its an Essick Air Tower Humidifier, though its about the same price as what my old Essick Air unit...though thinking the tower form would work better with where I've been using the old one, plus the old one's fan has become the issue... Though I guess its only 2 years old. I wish I could find something like the humidifier I had back when I was in Medicine Hat....while it eventually got nasty inside, that thing worked well for years and wasn't that expensive as I recall. And, it was easy to fill too. Since, then...I guess the Essick Air is the first one that I got more than one season out of....two whole ones, in fact (though some debate as to how well near the end.) And, it did a good job too. Plus the tower I'm looking at, uses the same wicks as the old one...so I already have an extra set somewhere.
Though having an extra set of wicks isn't a requirement for what I get...though footprint and output are...and pretty much I think wick types are the only kind I'll consider now. And, brand isn't necessarily a requirement...though past experience with other brands has made me leary of others. Plus there's feature differences. Some do take less space, but only have one tank. But, I'm running this single humidifier to try to humidify my entire condo...so it does have to be big as well... But, have time before I have to complete the order...so reviewing all the results....wouldn't be so painful if amazon's search was so fuzzy. I'm searching for just "humidifier", and its showing me things like cribs, pillows, swaddles, bassinets, dehumidifiers, evaporator coolers, air purifiers, bottle brush, nipple cleaner, heaters, bonsai tree, air conditioners...though the cigar related humidifiers probably aren't out of line, though I don't really think they should be in the department I'm searching under. And, the further out...finding humidifiers becomes less and less likely, but sometimes there'll be one here and there....though I'm sure if they weren't so fuzzy, I wouldn't have 10x the results to wade through....though sometimes the result is way near the end. But, not this time.
Anyways...that's pretty much what should just be in my amazon cart. But, if I add a few more items, I can get 6-month do not pay on my Amazon card....and I've already put off getting the humidifier too long already.
Pages: 1· 2
Latest Poopli Updaters -- http://lkc.me/poop
|<< <||> >>|
«amazon prime» tivo cox «doctor who» «windows xp» «tivo hd» «chicago tardis» b2evolution dvd virtualbox replaytv «sans digital» lhaven zen box «powersource 400» ups tardis freebsd raid usb «watch instantly» «instant streaming» linux mdadm netflix appletv tv backuppc prescription amazon.com eyeglasses wifi «tivo premiere» raid1 boinc cfengine3 «air purifier» progressive orac upgrade woot 10.04lts dsl twitter cpap ubuntu «windows 7» ebay «hd movie»