Because one of the 1TB RAID1 drives was starting to fail...I was looking at options to replace it.
I original had a plan to upgraded one of my 1.5TB RAID1's to a pair of 2TB drives....freeing one of the 1.5TB drives to test out the OCE/ORLM feature of the RR622 & TR5M-BP on Zen...converting the current 2x1.5TB RAID0 to a 3x1.5TB RAID5 (or perhaps 4x1.5TB) Oops!Backup seems to be doing well and its filling up the space, I'm sure there's a setting to tell it not to, but not an issue at the moment. Guess I should buy it before the 30 day trial runs out.
At first I was looking at new 1TB drives....though for $20 more I could get 2TB drives, the same kind as the 6x2TB RAID10. The 1TB and 1.5TB drives have been 7200RPM (as were all previous drives)....the Hitachi's are the first of the greener drives....5940RPM. Given that I'm doing PM RAID in a first gen PCI Express box, I'm probably still not seeing the full potential of the greener drives.
The other day when I was shopping for quieter 120mm fan for my TR5M-BP enclosure, I thought about getting a bigger/better fan for Orac. I knew there were holes to take a fan larger than the one that was current in it.
Though I guess I should've measured or something. Because after I had the old fan out and things apart...found the fan that I had gotten was too big. I started putting things back, when I decided I might as well do the other things I wanted to do.
First the mirror of the internal RAID1 was failing.....the original had failed early and I exchanged it for a recertified one from Seagate and put in a different drive. Interesting that the warranty of the replacement drive seems to have expired, while the warranty of the original ST3500630AS is still active.
I was trying to decide if I buy a new 500GB drive, or put the recertified one in. Seeing as I had things open...and I hadn't bought a new drive yet. I put the recertified one in. I also had that project to put in a PCI compact flash card in. Though I hadn't gotten the desired compact flash card, yet. But, did have a 16GB 133x card sitting around...so I decided use that for now. The 8GB 600x card will have to come later.
I then put everything back...and watched to take forever to boot....the PCI compact flash card seems to slow the boot down a lot, and I don't know why. Maybe I don't want it now. May need to rethink the experiment. Perhaps when I get a correct sized fan and open it up next.
And, then it didn't....guess in everything, I forgot to reconnect the case fan. Oh well, the silly clamp does a lousy job holding the unsupported slot covers in place...and putting cables back in, the bracket pushed in. They probably knew that since the blank has a hook on it, but the eSATA bracket is a regular kind...but no room to slip a screw in to properly secure it.
So, opened things up and then tried it all again.
Oh yeah, while I had it open the first time, I noticed the CPU heatsink was all full of dust, so I blew that out.
Well, things came up slowly....of course the boot RAID1 was degraded. I then sfdisk'd the new drive, and added it back to the md0/md1....should be done in an hour or so (from now, not when I added it...)
And, then went to fiddle with the compact flash card as an external journal of an ext4 filesystem. Forgot that I was originally playing around with the idea of external bitmaps when I had gotten the idea for the compact flash....but concluded that it wasn't going to work, there's no way to have a filesystem exist during that stage of boot. Not sure if an external journal helps with performance or not....might have been something back when I was doing RAID5/6....but not sure its important enough with RAID10. Don't really have any way to measure if it makes any real-world improvement to what I'm doing.
Meanwhile...I updated cacti, the card shifted all my drives around.... And, seems the CPU is now running a bit cooler than it was before....but the GPU is running hotter than before.
Guess clearing the dust out of the heatsink was good, but adding the PCI card was bad.
I think keep things cool is more important than the external journal experiment...so I will probably yank the compact flash card out when I get the new fan. Wonder if a PCI slot cooler would be something I put in instead?
This is an owed post (of over 3 months in the making....) I had thought of lots of things to write for this since I meant to write it, but didn't write anything. And, now its so old...that I'm just going to be really terse....
The lead in to this posting was Worked on Orac yesterday. So, I got those 5 2TB drives...and first tried to make it RAID5 using mdadm. And, started filling it up....though the performance didn't seem to be much better using XFS, plus the new backuppc seemed to have regressed on handling XFS.... so I heard about ZFS on Linux....as kernel modules that do the taint. And, tried out RAIDZ, with its variable striping to maybe to do better on performance. It seemed nice, but then stack dumps filling up my logs....rebooted and then did a zpool scrub, and it went to work fixing lots and lots of errors, and then a few unrecoverable errors.
Guess no ZFS, and the XFS didn't seem enough of a gain to make it different from everything else. Even though there's one really big problem with ext4. Resizing it causes corruption. Supposed to be fixed now, but before I switched to the new array, I grew the backuppc volume for 3.75TB to 3.8TB, offline...and it resulted in corruption still.
So, I was making the new backuppc volume the full size of the array. As I mulled over the situtation...I decided that I would try RAID10, sure that wouldn't get as huge a boost over the old RAID6 array of 1.5TB, but it would still be an increase...and no idea what I'm going to do when I need more space. But, first I had to get a 6th 2TB drive to do RAID10. Plus it would mean that I would need to do something else with the 1.5TB drives. One was going to have to come out. Sure it could run degraded initially (though did find that it would come back after a reboot, guess the boot making degraded arrays come online...only does it for OS filesystems. Though the boot get's stuck because it can't mount filesystems and the recovery of mounting just doesn't work still. Getting into single user is such a pain too. Though after I broke my sudoers file recently, I turned off hiddenmenu.
The hard part about making the RAID10, was figuring out the ordering of devs...so that it was mirror across the eSATA channels and stripping with in. There's only 2 channels, and 6 drives in this array...so figured that was the best way to go. Though now that I've realized that its only PCI express 1.0, not sure if that was the best way to go. Though to try the other, would call for me to get 8 drives? And, who knows what the future holds for supporting backuppc....
internal bitmap on doesn't seem to degrade the RAID10 as much as RAID5 or RAID6, so I have that turned on now.
Now there was the question of what to do with the 4 remaining drive bays and the 4 1.5TB drives. Would I got with another RAID10, a couple of RAID1...would I concat them into the LVM for growth or make them separate.
Well, on the old 1TB array was my old MMC volume, which needed more space and part of the 5x1.5TB RAID6 was a backup volume...also in need of more space. I decided that I would go with two separate RAID1s using the Seagate 1.5TB drives (I had pulled the Samsung 1.5TB to make it degraded....that drive has since been hooked up to TARDIS for local backups).
Around this time, my Roku appeared...and I went through the various attempts to map a network drive to it for local content. Settling on the HSTI Media Stick and its 1TB maximum. I settled on going with 2 1.5TB RAID1 arrays. One of them was made fully available to be the new MMC volume. And, the other...I made a 1TB volume (called TARDIS) for the HSTI Media Stick, and the rest became the new backup volume.
The old 1TB RAID1 array stayed online, until it was recently repurposed into the volume for Time Machine backups (only 931GB).
Now I've been debating playing around with external bitmaps and/or external journaling to see if I can get more performance. But, external bitmaps go away after a reboot, and that just seems to be a bug that isn't going to get fixed anytime soon. external journaling would require me to find some devices for that....so I got the idea of PCI Compact Flash adapter and some compact flash cards....I have a couple of the PCI Compact Flash adapters....and I do have a few compact flash cards around. Though when I was playing around with ReadyBoost on Zen, I found they weren't really that great on speed. And, I was going to want to get as much speed out of things as I could here....so I did some checking and now I'm waiting to get around to buying some 600x CF cards to get around to trying this. The fastest card I have available in my collection is a 133x. And, I don't recall why I bought that one. Ones earlier than that would be left overs from when I had my PowerShot S20, then an REB1200 and then a Nikon Coolpix 5700.
These days, I'm all about the SD cards for my digital cameras (Eye-Fi), though been thinking of getting an ultracompact (and wearing my holster less often)....and some that I've looked at use MicroSD cards. Don't have any extra Class 10 ones laying about though....though I do have a couple of Class 6 8GB cards that need a home (I got them in a 2for1 sale, and the need for 1 didn't happen as planned....and when I do get around to getting that device, I'm sure I'll probably cave and get a 16GB or 32GB Class 10 MicroSD card.)
So, during the long weekend...I got to wonder why I couldn't have my work MacBookPro do Time Machine backups to a network share, I had envisioned coming up with a drive to attach to one of my Airport Extremes and doing that some day. Though it doesn't look like that's actually going to happen anytime soon.
Did a quick google search on Time Machine and SMB, and found out about:
defaults write com.apple.systempreferences TMShowUnsupportedNetworkVolumes 1
And, so trying to make it work started. I created a Samba shared named 'SIDRAT' on ORAC...which I think is the name of the removable 160G drive that I had been using at home for Time Machine backups. The one at work is named 'TARDIS'. But, I already have a Samba shared named 'TARDIS' on ORAC....that's the 1TB volume for the HSTi Media Stick. Also not to be confused with computers named TARDIS and SIDRAT. Though SIDRAT died a long time ago, and I still haven't gotten around to replacing it....it was my Windows laptop.
At first I didn't have dedicated storage on ORAC for this...I was just exporting the subdirectory of another Samba share as SIDRAT. And, it didn't really make sense for it to be a subdirectory on that share...but it was where I had at least 80GB to spare to this....the drive in the MacBookPro is only 80GB.
It seemed to start, but then it would fail. I had picked just the usual options for creating the Samba share, and wondered if the Mac needed special stuff to work. So, I went looking....
Read about the:
trick...but that didn't work. And, wasn't necessary when I did get it working.
So, then I read about creating the sparsebundle to get it started. There was conflicting info on how to name it....the usual was start Time Machine, and see what name it creates and use that.
So, was creating
"<host>.tmp.sparsebundle" And, there were some variations on how to invoke hdiutil to create this. Well, it went further...it slowly made it get smaller and smaller until it was gone and then went on to try to create
"<host>.sparsebundle". Creating that as well didn't help. I had heard about
"<host>_<MAC_ADDRESS>.sparsebundle", but the hits I found didn't explain just what MAC_ADDRESS would look like, or which one to use. Decided this wasn't going anywhere and the long weekend was nearly over....I did at least (after many long attempts) get a backuppc backup of parts of my MacBookPro. So, I got rid of the Samba share and undid most of the other stuff.
It isn't that new....in fact I had picked it up from a clearance table a couple months ago. But, it sure is tiny compared to what was sitting in its place before.
For sometime, the latest storage configuration was a pair 300GB firewire drives and a 1TB firewire drive connected to the box and its 160GB system drive. The 1TB drive broken into two partitions...1/3 & 2/3...the larger as MMCVideo storage and the smaller for local stuff. The 1TB replaced what was 3 300GB firewire drives, when one of the drives failed....and it was a total loss, but I had restored some 400GB from backuppc to recover.
I had kept the 2 remaining 300GB drives as a two set disk2disk backup of the system drive (with the help of Retrospect 6).
Well, the remaining two 300GB drives had started failing a few months back. One of them failed completely a couple months ago....prompting me to eventually buy a new drive.... But, the other 300GB drive continued to work more or less and I changed the backup schedule to only use that drive. It did lead to some bad crashes now and then, but apparently not enough to take it off.
This morning while I was looking at other things, including tweaking backuppc...and push some fulls now that I'm in a quiet period of system activity (gumby is now more of a TiVo Desktop machine than MMC, but its kind of a quiet period for new TV...even though I'm really behind on my watching, and could find that some recording didn't happen and then need to find somewhere to buy the show....)
But, I decided it was time to retire the other 300GB drive before something worse happened...and break out the 250GB drive I had purchased.
It might be time I upgraded/replaced 'lhaven'.
This is the only computer in my home network that is only 10/100BaseT, everything else being gigabit. I did picked up a gigabit NIC, but couldn't get it to work in this computer. Seems to not get along with something integrated hardware-wise.
The other thing is there have been lots of errors coming from ata1.00 for sometime. This is also the remaining computer that uses PATA drives. 'lhaven' has 3 300GB PATA drives in it. The primary master (ST3300622A) and secondary master (6B300R0) were in RAID1 configuration, and the primary slave (WD3000JB) is just a single disk. secondary slave is the CD drive...forget what it is exactly.
It wasn't dire that I replace the failing primary master, though it had failed out once before....system booted off the secondary master and ran degraded. But, I added the disk back in and it quickly got happy again (have internal bitmaps enabled).
I had looked around and there isn't much in PATA drives available these days...and nothing in 300GB. Choices I found were 80, 320 or 500. I suppose I could've gone with a 320...but what I did instead of score a 300GB off of eBay.
Replacing it didn't go though. Failed out the disk, booted up on secondary, and tried to put in new disk. First cloning partition table over to new disk. Wouldn't work...new disk is smaller than the secondary master.
As it turned, out the two primary drives were the same size, but the secondary was slightly bigger. I had partitioned the drive using all the space, though it was never using all of it. Probably should've partitioned it the same size as the primary drives. But, I didn't recall this at midnight...figuring it was would go quick and let it resync while I slept....
So found details on shrinking things so that I could add the new disk. Then things got weird.
grub got messed up, and the new disk wouldn't come back correctly into raid. There seemed to be CHS issues, maybe the BIOS in 'lhaven' is too old for such a new drive (ST3300820A)...had seen messages about it, but was more stumped by why it kept changing the md1 slice from sda2 to sda...leading to nasty things as it tried to rebuild itself.
Also noticed that it did different things if it booted from the primary versus the secondary. Apparently the md0/md1 didn't show up soon enough so it used the partition directly.... and the sda1/sda2 paritions were kind of messed up. In fact the /home filesystem on one got seriously hosed, so I reformatted it. But, when I rebooted later..../home was back and working.
During this process grub got messed up, so I had to download a livecd to reinstall grub. Eventually, it all seemed to mostly be working. Except that I could only boot from the secondary master, it would come up wrong from the primary. And, the errors I was seeing that prompted this were still happening.
So, likely not a drive problem...but something else wrong with 'lhaven'. So, I decided to put the old primary drive back and resync that and put everything mostly back the way it was before.
It wasn't without some challenge...like how to make it not see the secondary master and get confused by it. I tried removing power to it, but computer wouldn't power up. I tried removing just IDE cable from it, BIOS halted because just a secondary slave wasn't kosher. It started things degraded, but used sdc1 as root (instead of md0 which should be sda1, with sdc1 out...zero'd it superblock) it was also weird on vg00. md1 wouldn't activate, but vg00 was using it some how. And, when I did activate it...I couldn't get sdc2 to add to it.
But, eventually it all got back to the way things were before.
So, I guess its time to see about replacing 'lhaven'. Its old enough. Though not sure what, besides case and powersupply are original from the computer I purchased in August 2002.
Hard to say what I'll replace it with though...since not much out there fits into where it is (perhaps its time to get a new KVM), and I'm going to want a pretty souped up system to speed up BOINC work....even though that's pretty much all it does these days. Since, everything is being handled on 'box'/'orac' now. And, the secondary services that do run on 'lhaven' don't call for anything beefy (DNS, NTP)....though I do want to redo my Linux media server setup.
My Momitsu V880N is probably going to get replaced sometime later this year... only had mmc running on 'tardis' and 'lhaven'. mmc doesn't work on vista/Win7, and 'tardis' is off in the corner to die. Could probably get mmc working on some other linux box if I needed, but its probably time to come up with something DLNA.... and have my Roku take over.
... an HP Pavilion Elite HPE-450t, customized with things like:
Should be good for doing BOINC
Now, I bought this system customized expecting to pretty much plug it in and go, only having to add externally attached peripherals to complete things. Though I do plan to do upgrades in the future, specifically I got the the 3 DIMM ( 2x4GB + 1GB ) configuration of slightly more memory so that I could potentially take it to the 16GB max at a future date without having to buy all 4 new DIMMs. If I had bought the 8GB configuration, it would've come as 4 2GB DIMMs. Anyways...without doing anything else first...I just had to take a peek inside.
Now, I don't know if it was that I got a Core i7-860 Dell computer at work (replacing a P4-2.8Ghz machine, so slower than TARDIS - a P4-3.2GHz w/HT machine) or ordering this computer...but TARDIS had seemed to get slower and slower, to pretty much insufferable these last few days. Still not ready to turn new system on....more later.
Was discussion storage speeds with some friends over lunch....so I wondered if I could get some data comparing the performance.
vg0 is are two internal SATA drives, each on their own SATA channel, being mirrored. vg1 is the SANS Digital pair of drives, sharing a SATA channel via Port Multiplication, being mirrored.
Guess the graphs show that the internal drives are faster for small files....though for larger the difference goes away, so that things are comparable.
As it is....vg1 is mainly my MMCVideo filesystem - MMC (Momitsu Media Center). So, the files tend to be large and sequentially read/written.
Meanwhile, I've already removed the old MMCVideo logical volume from vg0 and doubled the size of /home. Even though now that I think of it, there's stuff under /home that might better go in its own logical volume(s)....now that I have room to create them.
Or later go on that HP MediaSmart Server that I've been dreaming of adding to my network....
I've been wanting to try a Port Multiplier eSATA enclosure on my Ubuntu 8.04LTS Server, known as 'orac', for sometime. But, keep getting sidelined by how much it would cost to get a 5-bay enclosure and 5 drives... it would only be a matter of time that the price of drives would come down.
But, storage has been tight on my file server and it has been causing issues, and I know I'm going to have tons more stuff to add to it. The server current is just a pair of 500GB internal drives mirrored in RAID1. So, I started looking at options to add 1TB of RAID1 storage. I considered buying two discrete external 1TB drives....or buying an external enclosure that does RAID1. When I came across the SANS Digital TowerRAID TR2UT enclosure, and a few other like it....I narrowed on this one because it was under $100 and available under Amazon Prime. But, this enclosure sat in my cart 'saved for later'.
The reason I considered the enclosure was that it purports to do various RAID configurations on its own, additional it can be just JBOD and connect to a Port Multiplier port on 'orac'. If the drive didn't work on Port Multiplier, I could resort to the other options....additionally, it does USB...so if it just doesn't work on eSATA I could use it that way. Especially, should I later get 5 bay PM eSata boxes and displace this one. And, there's that thought that I might get an Airport Extreme and do TimeMachine backups to a USB attached storage....
Pages: 1· 2
On Monday there was a brief power outage....I didn't notice from work at first, because the UPS there were no issues affecting the computers at home that I accessed from work. Two didn't go down at all, and one rebooted by itself.
The Windows 2000 had started booting up, but since it mounts shares samba shares from the other box that didn't come up nicely...it was stuck until I told it to ignore trying to remount the share and continue.
The other box is the ubuntu server where the missing samba share originates from. I had set this system up as RAID1, and one of the drives stopped working after the outage. But, ubuntu wouldn't boot without both drives. Found I had to add
'bootdegraded=true' to the kernel line in grub.
Later the old APC Back-UPS Office 500 started making lots of noise. Time to replace its battery. Or perhaps time to retire it from service. I moved the loads to one of the remaining UPSs, and pulled the UPS while I investigated trade-in or replacement options.
Then this morning, I heard a strange continuous beep.....while trying to figure out where it was coming from....everything went dark. Or rather I saw my ubuntu server go down, and later on the KVM switch I saw that two other computers were booting up. Evidently, the remaining UPS was already handling too much...and its only a 700VA UPS.
I had forgotten that I had a plan to add a (Smart)UPS to the living room and move the Geeksquad 1285VA UPS eventually to run the ubuntu server (since that was the most recent addition, and it didn't have a UPS for it)....so it had been parked on the UPS that runs 'box'. Well, moving 'gumby' definitely put it over the top....probably especially with its 3 external drives. (used to be 4, but one died).
I guess it is not only time to rush on acquiring a UPS to replace the APC one that I retired, but also acquire a UPS to put the Linux servers 'box' and 'orac' onto separate UPSs.
I think I was eyeballing a CyberPower 1000VA UPS yesterday....the question is will it still go in the same place I was looking. Maybe I don't care that much about 'gumby'....though newer UPSs have lots of protected outlets, so I thought it might park nicely next to it and reduce the confusion of the which of the many powerstrips are plugged into which UPS, etc.
Am I at the point yet where a Smart-UPS can materialize, or maybe I'm looking at an additional CyberPower 1000VA (or perhaps something bigger, but still cheap)....
Latest Poopli Updaters -- http://lkc.me/poop
|<< <||> >>|
«doctor who» «air purifier» prescription eyeglasses ubuntu mdadm upgrade «instant streaming» usb voip woot cox dvd raid1 «chicago tardis» progressive cpap «watch instantly» backuppc lhaven «tivo premiere» zen freebsd netflix ebay «windows xp» «hd movie» tivo tardis b2evolution appletv «tivo hd» raid replaytv linux orac «powersource 400» virtualbox cfengine3 twitter boinc tv amazon.com 10.04lts box ups dsl «windows 7» «sans digital» «amazon prime»