I decided to add an SSD drive to one of my lab ESXi boxes; so I can utilize the Host Cache features, and test out the VMware VSAN beta (which requires SSD). I simply mounted an SSD drive into an HP caddy, booted into HP Offline Array Configuration Utility, made a RAID0 (with only one drive it is the only option), booted into VMware.
ESX did not recognize the drive as an SSD drive. There is a way to force ESX to categorize a drive as SSD, but do I really want to do that? To me if the system doesn't recognize it, there must be underlying issues.
http://pubs.vmware.com/vsphere-50/index.jsp?topic=%2Fcom.vmware.vsphere.storage.doc_50%2FGUID-99BB81AC-5342-45E5-BF67-8D43647FAD31.html
Turns out it is a bug in the Smart Array P400 Firmware when using Vmware. Also it was discovered that this RAID card when dealing w/ SATA devices will only transfer at 1.5Mbps, where SAS is 3Mgbs. Interesting, so will there be much of an improvement it the SSD drive is limited by the 1.5Mbps connection? Also the P400 does not have trim support.
So options:
-attach the SSD drive to the onboard SATA control off of the motherboard, and find a way to MacGyver a way to mount the drive inside of the case but not in the hot swap bay. Oh wait, there isn't any onboard SATA connectors!
-Buy a PCI/PCI-e SATA card, and MacGyver the drive mounting.
-Use an PCI-e SSD drive. Will a consumer grade card work with ESXi? Paying for an enterprise card on the VMware HCL for my lab use is not an option.
-Upgrade the P400 to a P410. Will the volume just be recognized? Second hand cards, with cables, and battery run over $150.
Increasing size of ESXi5.x datastores
Victim is an ESXi 5.1 running on a HP DL380 G5 with a Smart Array P400 RAID card. I migrated the 6 drive RAID 5 to 7 drive RAID 5. Logging into Virtual Center (vCenter Appliance v5.5) using the Vmware Vsphere Client. The configuation showed that the new space was now available. Normally one clicks on the "Increase" button, and add an extent.
http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1017662
For whatever reason, despite seeing the extra space, I was not able to add the extent.
I then used the Vsphere Client and logged directly into the host; from there I was able to add the extent and expand the datastore.
http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1017662
For whatever reason, despite seeing the extra space, I was not able to add the extent.
New Line of Western Digital Sentinels
Windows Storage server 2012 R2
Xeon CPU
DS5100 2.3ghz dual core, 8gb ram
DS6100 2.5ghz quad core, 16gb ram, comes w/ 2nd
AC supply
Upgradeable to ECC 32gb!!
Raid 0, 1, 5, & 10 options
Marvell RAID accelerator
Does have VGA ports & USB 3.0
http://www.youtube.com/user/WDBusinessStorage
Another interesting thing is that there is a SEPARATE 2.5" hard drive for the operating system on the 5100, the 6100 get two drive for the OS. This gives more usable space, and drive rebuilds won't take quite as long.
http://www.youtube.com/user/WDBusinessStorage
Another interesting thing is that there is a SEPARATE 2.5" hard drive for the operating system on the 5100, the 6100 get two drive for the OS. This gives more usable space, and drive rebuilds won't take quite as long.
Vmware VM migration withOUT shared storage~
When Vmware release ESXi 5.1, the ability to migrate VM's from one host to another without having shared storage! Enhanced vMotion/Enhanced Storage vMotion. So yes...two ESXi hosts with only local data store, the VM can be bounced back and forth. However, the VM must be powered down first. Microsoft has Vmware beat here, MS allows running machines to be migrated. Also one must have vCenter installed, configured, and running.
Don't have vCenter but got a project that requires to move a bunch of VM's? No problem! Download and configure the vCenter Server Appliance. A free 60 day trial is included. It is NOT necessary to join it to Active Directory.
Update; with the Web Client one can migrate VM's LIVE across local storage. It cannot be done with the thick/C# client. FWIW, it is greatly faster than Microsoft's Live Migration, at least thus far with out truly doing back to back testing.
Don't have vCenter but got a project that requires to move a bunch of VM's? No problem! Download and configure the vCenter Server Appliance. A free 60 day trial is included. It is NOT necessary to join it to Active Directory.
Update; with the Web Client one can migrate VM's LIVE across local storage. It cannot be done with the thick/C# client. FWIW, it is greatly faster than Microsoft's Live Migration, at least thus far with out truly doing back to back testing.
Windows 2012 NIC Teaming & Hyper-V
Recently I was repurposing some used servers. Much to my dismay, I am putting Windows 2012 on them for use in a Hyper-V cluster. First issue we noticed is that if you have Server 2012 R2, it cannot be managed by the 2012 Virtual Machine Manager. So instead of loading R2 on these boxes, we are loading R0.
Why are we loading Hyper-V? The answer is MONEY! Vmware has been jacking around there licensing lately, especially for those providers who offer hosting services. They actually charge by the gb of ram allocated! Then Microsoft, gives you free four VM's if you run their Hypervisor. So buy a copy of Windows Enterprise, get four free VMs, on VMware one would have to buy four copies of server. OUCH! Anyways I digress.
I really like the built in network card teaming feature in 2012. The only way I know how to get to it is by the Fisher Price Server Manager. To create a team just click on NIC Teaming, add the NICs, give it a name, choose the type of team one desires (FYI switch Independent seems to work without any configuration changes necessary to the switch.
After that is complete one may notice that the team's status should be a 2gbps connection. One may also notice that the individual NICS now only have the "Multiplexor Protocol" enabled and all the normal protocols such as TCP/IP have moved to the team.
After adding the Hyper-V role, the team gets striped from most of its protocols, a new one is added called 'Hyper-V Extensible Virtual Switch', and a new NIC shows up called VEthernet which has the protocols and IP configuration.
TIP! If you are going to team NICs together for use in Hyper-V and don't have a separate NIC for mgmt. Create the team first, then add in the Hyper-V role. I did it the other way around and had some very strange results.
Video card/monitor woes with high screen resolution
My bedroom PC needed to be retired, it was a Dell Precision 630; dual Xeon 2.8ghz CPU's w/ 4mb cache, 400mhz bus; a real screamer in its day. Alas blown capacitors on the motherboard started to make the PC slow, odd video issues, and occasionally not boot. Besides, I suppose it is time I finally let go of my last Windows XP machine. This machine has an ATI Radeon 256mb AGP video card. That card drives a Dell 27" (2550*1440 resolution) and an IBM 21" (1600*1200 resolution).
Its replacement is a Dell Precision T5400, Xeon L5420 2.5ghz Quad Core w/ 12mb cache, 16gb ram FB-5300. It is upgradeable to a 2nd cpu, and these "L" Series CPU's only require 50 watts of power vs. the normal 75~100 watts. It currently has a Nvidia GeForce 210 1gb ram PCI-E video card.
While setting up the T5400 it was on a different 21" monitor running at 1600*1200 resolution. Once moving this machine to its new home, I was greeted to see this:
Its replacement is a Dell Precision T5400, Xeon L5420 2.5ghz Quad Core w/ 12mb cache, 16gb ram FB-5300. It is upgradeable to a 2nd cpu, and these "L" Series CPU's only require 50 watts of power vs. the normal 75~100 watts. It currently has a Nvidia GeForce 210 1gb ram PCI-E video card.
While setting up the T5400 it was on a different 21" monitor running at 1600*1200 resolution. Once moving this machine to its new home, I was greeted to see this:
That machine booted up into Windows 7, and was just fine. Until I moved the screen resolution to its native 2550*1440, it then went pink and fuzzy again; dropping it to 1920*1080 it was fine. I knew the monitor was fine since it ran at full resolution on the old pc. I swapped out the DVI cable, same result. I swapped to the VGA connector (DB15), the picture was clear but maximum resolution was 2048*1150. Went back to DVI port on the video card, used a DVI->VGA adaptor, 2048*1150 was the result. Switched to HDMI, 1920*1080 resolution.
I tried five different video cards, and nearly the same results! Interestingly the ATI based cards show properly on post, only the Nvidia cards were fuzzy and pink at post. So I have ruled out the monitor, the dvi-cable, and the video cards as possible issues. It was starting to look like a flakey motherboard or something else strange.
A friend of mine mentioned that perhaps I need a dual link DVI cable. Huh? What is this dual-link you speak of?
http://www.tomshardware.com/reviews/tft-connection,931-8.html
http://en.wikipedia.org/wiki/Digital_Visual_InterfaceI acquired one of these Dual-Link DVI cables, and tadah! No more pink fuzzies and full resolution!!!
hard drive benchmarks.....v1.1
More not super scientific benchmarking....All test were on an HP ML350 G5, dual quad core, SmartArray P400 w/ 512mb ram, dual nics in multipathing mode, running VMware ESXi 5.1.
There are subtle differences in the three WD Sentinel tests; do be fair I only ran the test once, so the differences COULD be statistical noise.
This last one is running the benchmark directly one the Sentinel itself. Notice how much performance is being lost over the network. I would be curious to find out if this a result of how the network is configured, the NICs in the Sentinel, or just what can be expected with network traffic. By looking at these results, one could conclude that the 5400rpm drives in this machine are NOT the bottleneck.
Two 146gb drives in RAID 1
Six 600gb drives in RAID 5
These next three are all going to a Western Digital Sentinel RX4100; the two nics are bonded using LACP.
Raw Disk Mapping
Normal Windows File Share
VMDK
Cisco USC 240
Cisco "new" hardware.
It has two SD card slots, and the machines "BIOS" kind of sort of is located on the SD card, there is one blank partition on it, if one chooses to load an OS on it. The 2nd SD slot is for having a mirror copy.
Nearly all of the machine is configured via a web page. Including RAID card settings. I could not find the "silence alarm" button.
The machine has on board SATA/SAS controller, which is very limited. There is an option to upgrade it to a "mezzanine" card. The mezzanine card will only allow the use of 8 drives and is not supposed to be used with any other RAID cards. This particular customer changed their mind on the purpose of the machine to where it needs more drives. So Cisco UCS-9266 (made by LSI Logic) was ordered. For whatever reason Cisco was rather specific in that if it is in a 2 CPU system the card goes in slot 4, and slot 3 in a single CPU system.
This UCS-9266 card has two ports on it, the backplane has four ports on it. It will not recognize any drives beyond 12. There is other cables offered by Cisco a UCSC-6, UCSC-4, and UCSC-2; I was unable to discover what the differences between the cables is, nor find a picture. I am also unsure what the machine came with. Apparently Cisco ships these machines with one of two different back planes for the hard drives, and to use the UCS-9266 we needed the other one. Supposedly the UCS-9271 will work as it has 8 ports.
Another point of interest: when ordering, keep in mind if one is getting a 16 drive capable system or a 24 drive system. On the 16 drive system, the last 8 slots for hard drives do not have SAS/SATA connectors, despite there being drive blanks there. Cisco will not sell the parts to convert a 16 drive system into a 24 drive.
Western Digital Sentinel RX4100
The big brother to the RX4000. This one comes with 4gb of ram has the same processor. I didn't study it real closely but it appears to have the same motherboard as the DX4000, except this one has an actual VGA port, which the RX4000 lacks. It appears to have a propriety power supply, and the motherboard slot looks like a PCI-E card where the SATA connector board plugs into on both models.
Watchguard XTM22 issues & throughput speed
The patient is a Watchguard (WG) XTM 22; it is the middle of the bottom tier of firewalls offered. Some speed issues where being noticed on a Comcast High Speed connection with their "Boost" package.
The first part of trouble shooting was to find out what the connection is capable of; by directly connecting a PC to the Motorola Docis 3 modem.
After a bunch of pointless trouble shooting steps the modem was moved from Eth0 to Eth2. What is interesting about this is that on this family of WGs Eth0 & Eth1 are 100mbps connections while the rest are 1gbps. The cable modem also has a 1gbps connection, but since ones internet speed is no where near 100mb it should not matter.
The issues appears to be in the proxy engine of the firewall. Next a test with a filtered HTTP/HTTPs rule. A different test PC was used, so their might be a slight scew.
WG offers an AntiVirus scanning, called Gateway AntiVirus (GAV). It was turned off for this test.
Not much of a difference w/ GAV turned off. This time IPS was turned off.
IPS appears to be the biggest culprit. It should be noted that this is a SOHO firewall, it is meant for roughly 5 computers. Turn on all of the security features such as WebBlocker, GAV, IPS, ect. really takes a toll on the little ATOM cpu contained in this box.
One thing that WG has done in thier newer version is to have a "reputation defense" feature. WG maintains a hosted database of sites they consider both safe and harmfull. If this feature is turned on, everytime a computer goes to the website, a lookup is done. If the requested site is on the "safe" list, it bypasses all of the security features. Likewise if the site is considered "harmfull" the site is blocked, and no further scanning by the WG appliance is done. This frees up many cpu cycles on the WG firewall.
The first part of trouble shooting was to find out what the connection is capable of; by directly connecting a PC to the Motorola Docis 3 modem.
After a bunch of pointless trouble shooting steps the modem was moved from Eth0 to Eth2. What is interesting about this is that on this family of WGs Eth0 & Eth1 are 100mbps connections while the rest are 1gbps. The cable modem also has a 1gbps connection, but since ones internet speed is no where near 100mb it should not matter.
Great improvements! However WG could not provide an answer why this helped, nor why we are only getting less than half of the speed available. More trouble shooting!
First all of the fancy proxy features were turned off, no IPS, no AntiVirus, no WebBlocker, just straight up filtered rules any out going rule.
The issues appears to be in the proxy engine of the firewall. Next a test with a filtered HTTP/HTTPs rule. A different test PC was used, so their might be a slight scew.
WG offers an AntiVirus scanning, called Gateway AntiVirus (GAV). It was turned off for this test.
Not much of a difference w/ GAV turned off. This time IPS was turned off.
IPS appears to be the biggest culprit. It should be noted that this is a SOHO firewall, it is meant for roughly 5 computers. Turn on all of the security features such as WebBlocker, GAV, IPS, ect. really takes a toll on the little ATOM cpu contained in this box.
One thing that WG has done in thier newer version is to have a "reputation defense" feature. WG maintains a hosted database of sites they consider both safe and harmfull. If this feature is turned on, everytime a computer goes to the website, a lookup is done. If the requested site is on the "safe" list, it bypasses all of the security features. Likewise if the site is considered "harmfull" the site is blocked, and no further scanning by the WG appliance is done. This frees up many cpu cycles on the WG firewall.
Subscribe to:
Posts (Atom)