This post is more than 5 years old
1 Rookie
•
4 Posts
0
6254
August 10th, 2019 03:00
SCv2020 failover - ports are not automatically rebalanced
Hello,
we have SCv2020 connected with VMware cluster by iSCSI Software adapter and we have problems with port balance behavior if we perform storage failover tests.
Compellent have 2 controllers with Front-End 4 iSCSI 10Gb ports (2 ports per controller) and Back-End 4 SAS ports, all is connected.
Here is our connection list:
- ESXi host:
- iSCSI VLAN 1 - x.x.1.100
- iSCSI VLAN 2 - x.x.2.100
- SCv2020
- Fault Domain 1
- Top Controller x.x.1.31
- Bottom Controller x.x.1.32
- Fault Domain 2
- Top Controller x.x.2.31
- Bottom Controller x.x.2.32
- Fault Domain 1
Connection is from this official scheme:
We have 2 LUNs, LUN-1 is owned by the Top Controller, LUN-2 is owned by the Bottom Controller (automatically).
ESXi host see 2 paths to LUN (both VLANs).
And now the PROBLEM, if we gracefully restart any controller from DSM Client, LUNs are rebalanced to the active controller and everything is OK. If we restart any switch, everything is OK.
BUT if we disconnect both iSCSI cables from one controller, LUN stay on disconneted controller and is not automatically rebalanced to another controller which have active iSCSI connection. LUN on disconnected controller is inaccesible in the ESXi host. And I must rebalance ports manually in DSM.
So, what can I do with it? How can I set automatic port rebalancing if both cables on one controller are disconnected?
Or how can I set whole system?
Another question, we have active NBD ProSupport until 2020, but if I want create technical ticket on the web, it shows that we don't have support. So, for this technical questions we need have another type of support?
Thank you very much for your help.
Origin3k
4 Operator
•
2.3K Posts
0
September 29th, 2019 07:00
For my understanding your problem about pulling both iSCSI cables canot be solved. The SC faultdomain are for surviving a controller failover and not pulling both cables.
Question: do you use round robin as mpio policy?
As a Partner i sold more than 10K dell devices. If this was a new purchase its unlikely that your STAG isnt in the support db... but i have seen this if we bought Demo and Test devices for us.
If its a warranty renewal than its more common and i see it more than once You need to call the support team and have the order confirmation id from that renwal around. This is a second method to proof the identy of the unit.
Regards,
Joerg
DELL-Sam L
Moderator
•
7.6K Posts
0
August 12th, 2019 09:00
Hello DaveR33,
I am going to send you a private message and if I can get you to send me your controller serial# I can investigate your SCv2020 & see why your rebalance is not working. I can also investigate your system and check on your warranty status & see why it may not be showing correctly.
Please let us know if you have any other questions.
DaveR33
1 Rookie
•
4 Posts
0
September 29th, 2019 02:00
Hello Sam,
I wrote you more than 1 month ago, but without answer from you.
Can somebody help me, why port are not automatically reballanced please?
I can't use storage in production enviroment, because this is very critical issue.
Thank you very much for you help.
DaveR33
1 Rookie
•
4 Posts
0
September 30th, 2019 12:00
Hello Joerg,
thank you for answer, I wanted to simulate controller fail and pulling out ISCSi cables was probably only option, because I didn't want to try pulling out whole controller So, if the storage will reballance ports after complete controller fail, it's OK. But if I can't test it, I must hope, that it will works.
We have Lenovo storages and there I can do this failover test without problems, storages reballance all communication automatically. So this Dell behavior is strange for me.
We don't have renewal license, we moved our storage from a managed cloud of external company to our own private cloud and we don't have all informations.
Origin3k
4 Operator
•
2.3K Posts
0
September 30th, 2019 14:00
If you would like to test the CM failover.... just press the CM restart button within the GUI. Its the same as CM failure or more important to simulate a SCOS update which will also restart the CMs one after another.
We always test this and also document the behaviour.
Regards,
Joerg
rodrigo_bola
1 Rookie
•
11 Posts
0
October 2nd, 2019 15:00
hello Dave
So, what can I do with it? How can I set automatic port rebalancing if both cables on one controller are disconnected?
don't do automatic rebalance, the test you took is how you missed the both switches, so all front end are down
only do automatic rebalance with a controller reboot or switch reboot FC or Network.
thks
DaveR33
1 Rookie
•
4 Posts
0
October 3rd, 2019 05:00
Hello,
failover tests with restarting controller from GUI are OK.
When I restart controller, LUN which is active on this controller is in vShere for about 1 minute inaccessible, but VMs on this LUN are working normally without downtime (I tried run IOPS tests and everything was OK).
I have the newest firmware, so I can't try this update test.
Thank you very much for your time.
Best Regards,
David
Origin3k
4 Operator
•
2.3K Posts
0
October 4th, 2019 14:00
If your ESXi lost access than somehing isnt right. You should see a warning about path redundancy and most likely a "increased latency" message within the Task&Events.
Regards,
Joerg