
UNSOLVED
Can we have multiple gateways for one ScaleIO system?
We've installed a ScaleIO system with an IM and the system's working fine.
The system is correlated with Openstack via cinder driver through the gateway by REST API.
We know gateway HA clustering is available to avoid getting the gateway SPOF, but wondering if the following way is viable.
The Openstack has 5 clustered Controller nodes and all the nodes have a cinder driver.
We'd like a ScaleIO gateway to be installed into the each of the 5 nodes and to have the cinder drivers access to the local gateway.
If it is possible, we don't need a clustering service like pacemaker to manage HA clustered gateways.
By now we've installed gateways into the 5 nodes but have not been working.
The 5 gateways' installed with the command: GATEWAY_ADMIN_PASSWORD="XXX" rpm -i "PACKAGE" --nodeps.
And cinder.confs were fixed and restarted: san_ip ="LOCALOHST IP"
The log generated in api_operations.log: INFO audit - IP ADDRESS:/api/login
*We also replaced java-1.8.0-openjdk with jre-1.8.0.
Summarizing questions.
Can we have multiple gateways accessing to a single ScaleIO system?
If yes, what are we missing to make them work like credential issues?
We know it's not a common case, but any help will be appreciated.
Many thanks,
Seiji
Responses (7)
Solutions (0)
Hi Seiji,
Can you see what's in the "localhost_access_log.log"? What's the HTTP status code with "GET /api/login", is it 200 or 401?
You can try CURL or RESTED plugin for Firefox to log in. Please remember to use the MDM password, not gateway one.
Cheers,
Pawel
Reply Hi Pawel
In the "localhost_access_log.log" we can see the 2 logs below.
"GET /api/login HTTP/1.1" 428 -
"GET /api/types/Domain/instances/getByName::default HTTP/1.1" 401 -
I tried a CURL command below.
curl -k -v --basic --user admin: https:// /api/login
and the response has
--------------
HTTP/1.1 428 Precondition Required
...
Connection #0 to host 10.X.X.X left intact {"message":"","httpStatusCode":428,"errorCode":0,"certificate":{"Serial Number":"1","Subject":"OU=ASD, O=EMC, C=US, ST=Massachusetts, L=Hopkinton, CN=XXX, GIVENNAME=MDM","Signature Algorithm":"SHA1withRSA","Signature":"[XXX","Issuer":"OU=ASD, O=EMC, C=US, ST=Massachusetts, L=Hopkinton, CN=XXX, GIVENNAME=MDM","Valid-From":"Tue Sep 06 20:14:10 JST 2016","Valid-To":"Sat Sep 05 21:14:10 JST 2026","Key-Usage":"Not Available","Public Key":"[XXX","Thumbprint Algorithm":"SHA-1","Thumbprint":"XXX"},"typeOfError":"SERVER_CERTIFICATE_NOT_TRUSTED"}
--------------
We are sure to use the right password, but whatever words we use as for the password the response is same.
So this looks like an SSL cert issue before password authentication.
When we try the CURL command using the gateway with that we installed the ScaleIO system,
we get a key(token) in the HTTP response.
How can we have the SERVER_CERTIFICATE accepted?
Best,
Seiji
Reply Hi Powel,
The SSL cert issue has been solved.
We copied the truststore,jks from the gateway with which installed the ScaleIO to the 5 new gateways and restarted them.
Now that the CURL command get a right response and cinder driver can create volumes in the ScaleIO.
*We didn't try the IM Web GUI because the Openstack controller nodes don't have GUI environment by dault.
Anyways you led us up to here.
Thank you very much.
Seiji
Reply Anonymous User
2 Posts
1156
1
Posted January 8th, 2017 22:00
Hi Pawel,
I faced similar problem in VIPR->SCALEIO integration and was able to resolve it after following the steps provided by you.
Thanks a Lot
Reply

smatsuda
14 Posts
1156
0
Posted September 26th, 2016 18:00
We modified gatewayUser,properties like below.
But it is still not successful to access the ScaleIO system through the newly added gateways by REST API.
--------------
mdm.ip.addresses=10.X.X.X;10.X.X.X *added Master and Slave MDM IP address
features.enable_IM=false *to use only for REST API gateway, We don't think this is relevant to the issue.
system.id= *left empty since the ID will be created on the first login.
--------------
After fails of attempts to create a volume via cinder driver, system.id has been left empty.
We get packets to and from Master MDM IP:6611. So the first login fails.
We also added system.id retrieved from the gateway which originally installed the ScaleIO system.
Just api_operations.log keeps getting nothing but "INFO audit - IP ADDRESS:/api/login".
Any ideas?
Many thanks
Seiji