So, I looked for the previous conversation I had created on this, but I couldn't find it. I am assuming it was purged as "unanswered" or just fell off the list because of age.
Last year I started a conversation that I was unable to respond to because of life, work, etc. Ultimately, the testing I needed to be able to do required more time and hardware then I had to give. In short, I run virtual machine labs via VMware Workstation Pro. I virtualize everything from your everyday Windows desktop OS to Server, Linux, and OS X. From a hardware perspective I'm using a Rampage IV Extreme with 64gb of RAM and a 6-core i7 processor. I have installed a 4-port PCI-e SATA III card for additional drives and also had an external 8-bay USB3(HASP)/eSATA external enclosure from Startech. Storage is as follows: 1 Samsung 850 Pro 512GB SSD for Operating System, 1 Samsung 850 Pro 1TB SSD for gaming, 8 internal drives of random size, speed, and form factor, but are all mechanical drives. The 8-bay external holds 8 additional mechanical drives, also of random size, speed, and form factor.
All 16 (non-SSD) mechanical drives are joined to a single pool drive. Duplication is enabled for archive data folders and I made sure that guest VM folders were NOT included. At one time I enabled duplication on the guest VM folders at 2x, 3x, and 4x to try and take advantage of higher read speeds. I found this more troublesome for guest VM's, and got more performance from migrating to split disks within the VM infrastructure for high use virtual disks. I use the balance plugin so if I have a 20gb drive split into 4gb files, then the virtual disk is essentially split 5 ways. I am still only reading on a single disk, but as the VM's grow read data is split more and performance is still better than duplication read speeds from the pool. This has been my experience and could be related to the issue below.
The issue I had/have is that when I start more than 2 "actions", i.e. file copy, vm guest etc., each of them will randomly lock up and I will only see activity on no more than 2 of these actions at a time. During this time most of the actions are "not responding", performance tanks, and I can see wait times as long as 20-30 minutes before one that was "working" goes not responding, and one of the other tasks starts "working" again.
The information requested at the time was basically about eliminating or isolating different hardware and configurations in order to see if this issue continued. Since that time I have done the following:
1) Removed the external 8-bay from the pool. NO CHANGE IN "NOT RESPONDING"
2) Removed all duplication from the pool. NO CHANGE IN "NOT RESPONDING"
3) Removed all drives in the pool but 1, a WD Black 2TB 7200 RPM drive. NO CHANGE IN "NOT RESPONDING"
4) Added a 4-port 1gb server NIC and have 5 teamed 1gb ports and configured LACP on my managed switch to see if adding sessiosn would help. NO CHANGE IN "NOT RESPONDING"
5) Wiped and rebuilt OS (Windows 7 x64) and tested each of the above again. NO CHANGE IN "NOT RESPONDING"
6) Added a Samsung 840 Pro 500GB SSD and isolated all disk activity here (guest VM's, file copies, etc) and had NO ISSUE. Never saw "not responding in this format". This included 4 threads from an external FTP (work PC), 11 guest VM's, which included 2 FreeNAS virtual machines with 6 virtual disks each, a DC, a WSUS server, 2 Hyper-V 2012 R2, 2 Hyper-V 2016, 2 ESX 6.5, and a management server with SQL 2016, vCenter Server, Hyper-V, and SCVMM, along with two separate 5gb ISO copies. I was absolutely KILLING this drive and it never hiccuped a single time. On the pool, I can only do a combination of any 2 of those.
To this day I still have this issue and am at the point where either I am using DrivePool in a way that it was never intended or capable of being used, or I have an issue on my system. Hardware was all tested "good", all drivers, Windows updates, etc. up to date and everything operates fine.
It seems that there is a maximum thread count on the DrivePool driver or in how the system handles activity on that pool/infrastructure. This issue completely disappears when I move all activity away from the DrivePool pool.
Thoughts, ideas? At this point, my only option seems to be to try something like DriveBender to see if it has the same performance hit. Perhaps there's a "secret setting" that a dev knows about that would fix this for me. I completely understand that I may be the .000001% of the use cases that use the product in this way. I may have this problem no matter what I do, but I'd rather not have to go back to splitting all my files and VM's across disks manually.
Question
zim2323
So, I looked for the previous conversation I had created on this, but I couldn't find it. I am assuming it was purged as "unanswered" or just fell off the list because of age.
Last year I started a conversation that I was unable to respond to because of life, work, etc. Ultimately, the testing I needed to be able to do required more time and hardware then I had to give. In short, I run virtual machine labs via VMware Workstation Pro. I virtualize everything from your everyday Windows desktop OS to Server, Linux, and OS X. From a hardware perspective I'm using a Rampage IV Extreme with 64gb of RAM and a 6-core i7 processor. I have installed a 4-port PCI-e SATA III card for additional drives and also had an external 8-bay USB3(HASP)/eSATA external enclosure from Startech. Storage is as follows: 1 Samsung 850 Pro 512GB SSD for Operating System, 1 Samsung 850 Pro 1TB SSD for gaming, 8 internal drives of random size, speed, and form factor, but are all mechanical drives. The 8-bay external holds 8 additional mechanical drives, also of random size, speed, and form factor.
All 16 (non-SSD) mechanical drives are joined to a single pool drive. Duplication is enabled for archive data folders and I made sure that guest VM folders were NOT included. At one time I enabled duplication on the guest VM folders at 2x, 3x, and 4x to try and take advantage of higher read speeds. I found this more troublesome for guest VM's, and got more performance from migrating to split disks within the VM infrastructure for high use virtual disks. I use the balance plugin so if I have a 20gb drive split into 4gb files, then the virtual disk is essentially split 5 ways. I am still only reading on a single disk, but as the VM's grow read data is split more and performance is still better than duplication read speeds from the pool. This has been my experience and could be related to the issue below.
The issue I had/have is that when I start more than 2 "actions", i.e. file copy, vm guest etc., each of them will randomly lock up and I will only see activity on no more than 2 of these actions at a time. During this time most of the actions are "not responding", performance tanks, and I can see wait times as long as 20-30 minutes before one that was "working" goes not responding, and one of the other tasks starts "working" again.
The information requested at the time was basically about eliminating or isolating different hardware and configurations in order to see if this issue continued. Since that time I have done the following:
1) Removed the external 8-bay from the pool. NO CHANGE IN "NOT RESPONDING"
2) Removed all duplication from the pool. NO CHANGE IN "NOT RESPONDING"
3) Removed all drives in the pool but 1, a WD Black 2TB 7200 RPM drive. NO CHANGE IN "NOT RESPONDING"
4) Added a 4-port 1gb server NIC and have 5 teamed 1gb ports and configured LACP on my managed switch to see if adding sessiosn would help. NO CHANGE IN "NOT RESPONDING"
5) Wiped and rebuilt OS (Windows 7 x64) and tested each of the above again. NO CHANGE IN "NOT RESPONDING"
6) Added a Samsung 840 Pro 500GB SSD and isolated all disk activity here (guest VM's, file copies, etc) and had NO ISSUE. Never saw "not responding in this format". This included 4 threads from an external FTP (work PC), 11 guest VM's, which included 2 FreeNAS virtual machines with 6 virtual disks each, a DC, a WSUS server, 2 Hyper-V 2012 R2, 2 Hyper-V 2016, 2 ESX 6.5, and a management server with SQL 2016, vCenter Server, Hyper-V, and SCVMM, along with two separate 5gb ISO copies. I was absolutely KILLING this drive and it never hiccuped a single time. On the pool, I can only do a combination of any 2 of those.
To this day I still have this issue and am at the point where either I am using DrivePool in a way that it was never intended or capable of being used, or I have an issue on my system. Hardware was all tested "good", all drivers, Windows updates, etc. up to date and everything operates fine.
It seems that there is a maximum thread count on the DrivePool driver or in how the system handles activity on that pool/infrastructure. This issue completely disappears when I move all activity away from the DrivePool pool.
Thoughts, ideas? At this point, my only option seems to be to try something like DriveBender to see if it has the same performance hit. Perhaps there's a "secret setting" that a dev knows about that would fix this for me. I completely understand that I may be the .000001% of the use cases that use the product in this way. I may have this problem no matter what I do, but I'd rather not have to go back to splitting all my files and VM's across disks manually.
Thanks for your time and any help you can give!
Chris
Link to comment
Share on other sites
13 answers to this question
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.