v2 .deb gateway installer really broke

Hi Hinch,

The included documentation covers the required parameters for running the Gateway setup on Ubuntu on page 64 of the Deployment Guide. I noticed your install command did not follow this syntax. Also, it seems that Ubuntu installation needs a good bit more manual work for the deployment through the Installation Manager (driver.conf file setup and other things listed in the deployment guide).

Aliasing sysctl to a systemctl command doesn't work because they are very different programs. systemctl is part of the systemd init environment which isn't fully installed by default on 14.04 (shipped with Upstart init manager instead).

I may poke around with the deb installer as well on 14.04 and 14.10 to see what else is off.

FWIW, I have had no issues deploying v2 on RHEL 7 yesterday. Hopefully we can get the Ubuntu support off the ground.

Cheers!

I used the quick start guide linux pdf from the downloads section and copied the commands from there (its the 13 page pdf not the 202 page one) if they're different from the full 202 page deployment guide then that'll be one of the issues.

The other issue is the the quick start says something like "dpkg -U /path/scaleio_gateway.deb" etc which is wrong as dpkg doesn't have a -U flag that is rpm. it should be dpkg -i

As for the alias not working your init script simply calls "systemctl start" which alias's perfectly over "sysctl start" they both provide the same functionality so I assumed it would work for that call clearly it didn't.

I've given up on the ubuntu for now as its just a nightmare to even attempt to make the installer work at current, have swapped over the 3 test nodes to centos 7 and its gone on fine fully up and running now I'm just trying to wkr out where the --map_volume_to_scsi_initiator has vanished to in v2 as it appears to be no longer valid

BTW, I don't work for EMC. I'm just another interested user

I downloaded Ubuntu 14.04 Server and loaded it up in a VM with only the SSH server bundle enabled for install.

After install, I added numactl, bash-completion, and libaio1 packages to the installation.

I then added the Oracle Java 8 runtime by using the webupd8team ppa and installing it using the Oracle installer (since openJDK 8 is not available for 14.04).

I ran the gateway installation using the recommended syntax from the Deployment Guide (pg 64). However, like you mentioned, the Tomcat instance did not listen on IPv4, just IPv6. A quick google showed this is an issue with Java's default preference toward using IPv6 if it's available. To adjust the Java preferences, I edited the file at /opt/emc/scaleio/gateway/bin/setenv.sh and added two more parameters to the line for CATALINA_OPTS=

The two added parameters were "-Djava.net.preferIPv4Stack=true -Djava.net.preferIPv4Addresses=true". After this I restarted the ScaleIO Gateway using "service scaleio-gateway restart" and confirmed it is now listening on IPv4 addresses. I was able to browse to the Gateway interface and move forward with the installation process.

I agree this should be noted in the deployment guide or fixed as part of the install. I did also get the systemctl error and the rpm error but those didn't seem to cause any problems.

Hope this helps!

hinch,

Binutils looks to be the main issue.  A fresh/clean install of the gateway would likely succeed.

I just tested this on a fresh Ubuntu Server 14.04.3 install (with binutils present) and despite the complaints of rpm and systemctl not being found, the web installer was running and accessible.

I think there's a whole host of issues that really need sorting and certainly the documentation needs improving.  It would also be nice if I could get a definite answer as to what has happened to iscsi in version 2 as they seem to have been completely removed from what I can tell but with no documentation regarding it.

Interesting that the setenv changes worked for you I attempted the same thing and it didn't make much difference.

dpkg -U is indeed a typo,thank you for noticing, we will correct it.

I don't see that you used the 'GATEWAY_ADMIN_PASSWORD' parameter as requested -> we re-tested it and it's not what's causing the failures you see.

Seems like the binutils is indeed mandatory. How come it was missing on your server? I thought it comes with Ubuntu 14.04 out of the box...

Taken from the Quick Start Guide:

Run the following command (all on one line): GATEWAY_ADMIN_PASSWORD= dpkg -U /tmp/ EMC-ScaleIO-gateway-2.0-XXX.X.deb, where new_GW_admin_password is a password that you define to access the IM.

As for the iSCSI commands they are no longer supported, even before 2.0, since we stopped supported running the SDC on ScaleIO VMs (SDC runs on the ESXi host since v1.32).

the systemctl error printout is a redundant error that can be ignored (not bad impact here). This issue will be fixed as part of 2.0.0.1 (v2.0 Patch1).

I just copied a different attempt to install line thats all I tried setting the parameter as in the quick start chained through I also tried setting it as an export prior to run then just running dkpg stand alone (this is the version I copied)  makes no difference.

No bin utils doesn't get installed out of the box on 14.04 or 15.xx server.

If you've removed iscsi and moved it in to a plugin for esxi then how are we supposed to use scaleio as general storage or hook it in to a different hypervisor  that isn't esxi.  For example I use xenserver and KVM/QEMU without iscsi it I would assume I can no longer use scaleio ?

ScaleIO 2.0 can be installed on XEN 6.5 server (tested and approved) and there's no need for iSCSI ScaleIO commands.

Please explain...

hinch,

Looking deeper, I see that binutils was not installed by default on 14.04.3, but as a pre-dependency of oracle-java8-installer. (Found at the webupd8 PPA as mentioned above.)

I'll pass this info along and make sure it's addressed.

hinch,

To use ScaleIO with Xen or KVM, etc., you install the ScaleIO components in the hypervisor's OS and map volumes to the hypervisor's OS SDCs. These volumes will show up as /dev/scinia, /dev/scinib, etc, which you can then configure in Xen.

Don't suppose you've got a brief guide / step by step for this or at least an architecture map of how this all works?

I'm struggling to get my head around how it all works together simple things like how to install sdc only on a hypervisor os.

EDIT:

Just been reading the User Guide and from my basic understanding its describing running scaleio within the actual hypervisors.  I'm running scaleio on centos on bare metal the hypervisors are on different machines as I didn't want the overhead of running the storage in vm's also.

Little update:

Yay I got SDC's running on both xen hosts, after much fighting with sio not accepting multiple sdc mapping of volumes via command line (works fine via GUI though!) I got the SR setup in my xen pool for the test nodes.

Got a test vm installed and running off the SR just fine.  Only problem I seem to have now is HA isn't working xen won't let me configure HA as it seems to think that the SR doesn't support it not sure why at the moment am open to any suggestions though.

So you just need to install the ScaleIO SDC on your XEN hypervisors (there are specific artifacts for XEN 6.5), either by using the Installation Manager, or manually.

All options are described in the v2.0 Deployment guide / User guide.

Hi,

I refer you to pages 455-456 in the User Guide -> Headline: Xen v6.5 High Availability

Note that steps 2-5 might be redundant in v2.0, but no harm in checking the rules exist with the proper configuration.

Thanks,

Tomer