Question

Errors on install - autofsb and drive number


Badge +5
hello.
I have an older box i'm trying to load CE on. I tried this before - but work got too busy and haven't had a chance to revisit - which i'm trying to do again now.

I have the following config - 32gb ram, quad core, 8gb usb (boot), 1 200gb flash drive, and 2 4-tb hard drives. somehow I created a bootable usb earlier but can't find notes on creating that again.

The drives are seen in the bios but I'm getting a 'autofsb' error error before I try to install. there used to be data on these drives - but I don't need it and can delete if someone advises how.

question are the following:
  1. how do I create another bootable usb drive with the newer CE software (downloaded this am)? Not that this will fix - but will mean the i'm current in testing. I have Rufus and have used it before to create bootable media.
  2. do I have enough drives - or need 1 more? doc's say 4 max but don't see if that the minimum also. I don't have another 4tb drive, but could 'mix' a smaller 1 in if needed.
  3. how do I get around the 'autofsb' error? I presume that's the program determining the hardware as when ever the 'install' tries to work - I get another error about # of drives required.
I have 3 identical hardware boxes (quadcore, 32gb ram, 250g flash) that i'd like to add additional drives to and make a cluster - if I can grasp the install process.



thanks in advance -
mark

This topic has been closed for comments

22 replies

Badge +5
forgot boot image of drives -

Userlevel 7
Badge +34
Thanks for sharing @seapro220

I've moved this post into the correct CE forum. I'm hoping others here can help

--
Also consider checking out the I'm new here, how do I get started? post as well as the How to Change Your Community Profile Avatar - These are some great starter posts for getting a sense of the community.

cc: @andrew_ct @Primzy
Userlevel 7
Badge +25
I would follow this post by Adam and post the results

https://next.nutanix.com/discussion-forum-14/nutanix-install-on-hpe-proliant-dl360-gen10-a-set-of-disks-matchinf-the-minimum-requirements-was-not-found-27515/index1.html#post28087

The posts that @aluciani puts in will help with reimaging your USB if you want to.
Badge +5
guess I missed the creation process. I downloaded/extracted file (ce-2018.01.31-stable.img) and tried to 'rufus' it - got an error about image being corrupt. trying download again. rufus didn't recognize the file as a 'iso' - if I remember correctly, but it did change the bootable-image to 'dd'. isn't it correct that I just point - whatever the file name to - and rufus will create the bootable image, on the usb?

once I get past this point - i'll run fdisk -l and see what's going on with the drives. 250 gb is small and near the minimum - but should work I think to test with, correct?
Userlevel 7
Badge +25
Yeah you need to use the DD style as it isn't an ISO, but a copy of the filesystem and rufus should detect that.

Use the python code I linked to as well as that is how nutanix is classifying the devices and will help spotting issues.
Badge +5
While I wait on rufus - i'm sure that 1 of the drives had some previous data on it and that might be the reasoning that CE is complaining. while in as root - is there a fdisk command whereby I can delete the/any data not needed on that drive so it's 'clean'? would fdisk /dev/sda Drive# work?
Badge +5
running the fdisk -l command shows 4 drives.
/dev/sda is a 250g that I want to remove all data from it.
/dev/sdb is my 4tb drive
/dev/sdc is my 8gb boot drive
/dev/sdd is my other 4tb drive.

after I delete all the data from it - fdisk d 1,2,3 do I need to create another partition - formatted in any specific manner? I deleted the partitions, exited out - in as 'install' but am getting the same error.
Userlevel 7
Badge +34
Just a side note @seapro220

I use Etcher instead of Rufus - simple and easy to use.
Badge +5
running the ./collect_disk_info.py command reveals the following info. not sure I know 'all' that it gives, but somebody must. I suspect that the install might be faulting due to the 250gb drive is a sata - and not flash but not sure. if that is correct, and I can modify the /home/install/phx_iso/phoenix/minimum_reqs.py command to change it - what editor do I use or how do I modify it? is there an onboard editor within the usb-cd-image somewhere?
thanks,

Userlevel 7
Badge +25
You can use vi as that is available to edit. Google can help with navigation as it takes a few minutes to get your sea legs.

So looks like you have 2x Seagate Constellation 4TBs and a 1x Seagate Momentus 250GB HDDs. Even if you hack the installer I am guessing your cluster will have major issues if it starts as all. The HDD is 1000x slower media and Cassandra likely will fail in nasty ways trying to run off a 100 IOP spindle.

The minreq is not there to drive demand for flash. 😉 To run CE you will need to pickup some sort of SATA flash SSD drive in the 200+GB range.
Badge +5
AArrggg !! was afraid of that - but was hoping to just get "1" running to see/play with the interface. the sata/flash drive I know I have is only 120gb which isn't large enough. w/o going crazy and spending a lot of $$ - is there a sweet spot of 200+gb drives out there that anyone's using? i'm probably going to have to buy some additional (3 per host) to get the cluster up - but those could be in sata drives cheaply - and not sata/flash.
Userlevel 7
Badge +25
May want to give these a go https://portal.nutanix.com/#/page/docs/details?targetId=Nutanix-Community-Edition-Getting-Started:Nutanix-Community-Edition-Getting-Started

Some flash is better than no flash so maybe that will work if you hack minreqs to allow for 100GB. I don't know if the Java heap for cassandra and other hard partition sizes might just kill that, but it will be tight.

So some folks have had issues with some flash (samsung's EVO family oddly has had issues) so I would avoid the bottom of the barrel if you can. Figure 75-100/SSD and shop your sales. Enterprise class doesn't hurt, but adds significant cost. I don't know of a specific list of validated consumer SSDs yet. @aluciani or @AdamFG Pulse has some data on hardware so could that be used to try and build a generic heat map for hardware being used in CE? R0s likely will create issues, but something is better than nothing?

For a demo you might want to try https://www.nutanix.com/test-drive-hyperconverged-infrastructure/ which runs in Ravello (last I checked).
Badge +5
Ok - i'll shop some SSD/Flash drives and see how the rates stack up. I think the 250GB Seagate is causing the issue as I thought it was SSD until you pointed it out - and I removed and looked at it. If I remember correctly I only need 3 drives - 1x200+ ssd, and 2x-sata drives for files (not mentioning the usb-boot device). is that correct? 4-drives are the max but I can use only 3 if I meet those requirements right?
Userlevel 7
Badge +25
You only need 1x SSD at 200GB+ and 1xHDD at 500GB+. Technically you need a 8GB+ other block device, but that's the USB. Only 3 devices is essential with one being a 200GB SSD as the metadata drive where the important stuff runs.

You can have more HDDs for capacity
You can have a 500GB+ SATA SSD instead of an HDD for capacity
You can use NVMe as long as you have another SATA connected device (install logic defect)

You can also technically have more than 4 devices (I had 9 [2 SSD and 7 HDDs] mapped and 1 HDD left for the rootfs after burning through too many flash disks), but understand the way that CE avoids a dependence on hardware that allows for device pass-through is it uses virtio as an abstraction for the CVM. That means there is a single IO queue so as you add more devices they won't necessarily make things go faster (and can actually slow things due to context switching) as they all need to go through a single tollbooth. So 4 is what the nutants tested and feel comfortable virtio won't implode. 😁
Badge +5
Thanks for the clarity. As I have 3 4-tb (1 not working for some reason) I should be able to obtain 3 200gb-flash and get the cluster up. I'll shop for some drives and try to get 1 working - then worry about additional drives for the cluster.
Userlevel 7
Badge +25
Yup ce has an add_disk script that usually works. 😅
Userlevel 7
Badge +34
@seapro220 As you work through this - If you find an answer that helps, please consider clicking the 'like' and 'best answer' link as that will help others in the community find the answer much quicker - Thanks
Userlevel 7
Badge +34
How are things going @seapro220
Badge +5
Got my 250gb and drive today so I'll be playingt tonight, I hope, got a wd-blue 250ssd. Should be ok I hope
Badge +5
Finally got a chance to start the config. gave a 'local' ip/subnet and a CVM address. I chose to do this because i'd like to have a cluster - down the road, but not at this point. the system is waiting for the CVM to start up - which obsiviously won't as there isn't 1 at this point. did I make the wrong choice - and enter a CVM address? both addresses answer to ping now but can't access the device - from ip addresses. I can putty to the CVM address but not the ip of the box and neither answer via browser (80 or 443).
Badge +5
Figured it out. now to start with some VM's and see how it/well-it works. I have a vsphere environment and utilize veeam as my backup solution - with replication. i'd like to try to build some 2016 servers - use veeam to replicate to and do some 'conversions' between vsphere and Acropolis - so my servers are running 'pure' acropolis. once I get the 'feel' of the environment I have 2 other identical machines that i can cluster together.

i'll look over some more documentation and watch some video's. any suggestions other than 'here' and what I've mentioned?

mark
Userlevel 7
Badge +34
Thanks for sharing @seapro220

Seems like you are on your way to becoming an expert!

If one of the answers posted helps (even if its yours), please consider clicking the 'like' and 'best answer' link as that will help others in the community find the answer much quicker - Thanks for your contributions.