Bug 1388566 - scanning storage interconnects and multipath
Summary: scanning storage interconnects and multipath
Keywords:
Status: CLOSED INSUFFICIENT_DATA
Alias: None
Product: Red Hat Enterprise Linux 7
Classification: Red Hat
Component: doc-Storage_Administration_Guide
Version: 7.0
Hardware: Unspecified
OS: Unspecified
high
unspecified
Target Milestone: rc
: ---
Assignee: Marek Suchánek
QA Contact: ecs-bugs
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2016-10-25 16:36 UTC by Gopinath Marappan
Modified: 2019-03-06 01:05 UTC (History)
4 users (show)

Fixed In Version:
Doc Type: If docs needed, set a value
Doc Text:
Clone Of:
Environment:
Last Closed: 2017-10-17 16:37:51 UTC
Target Upstream Version:


Attachments (Terms of Use)

Description Gopinath Marappan 2016-10-25 16:36:54 UTC
Document URL: 

https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_Linux/7/html/Storage_Administration_Guide/scanning-storage-interconnects.html

https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_Linux/5/html/Online_Storage_Reconfiguration_Guide/scanning-storage-interconnects.html

According to the above links, it is recommended that all IOs on the affected interconnects be paused and flushed before scanning the bus.  Else IOs might fail.

Does this also apply in the following scenario:
1.  Server is connected to a storage array through multiple buses.
2.  Only one bus is scanned at a time.
3.  dm-multipath is active and managing the paths.

In the above scenario when one bus is being scanned, if it takes a lot of time for the IO to complete on that bus, will it be redirected automatically through other available buses instead of failing that IO by dm-multipath.

Also, are there are any recommended parameters for multipath to ensure that IOs wont fail when individual buses are scanned, one at a time.


The reason for this bugzilla is to get some clarity on this issue because in our environment we need to add new LUNs from a storage array (which requires the buses to be scanned) while IOs are going to existing LUNs.  We dont want to pause IOs to existing LUNs.

Comment 1 Gopinath Marappan 2016-10-25 16:58:18 UTC
Specifically we want to know if the following command will require the IOs to be paused in both scenarios - without and with dm-multipath active.

echo "- - -" > /sys/class/scsi_host/hosth/scan

Comment 3 Steven J. Levine 2016-10-25 18:12:03 UTC
Reassigning to correct doc component for this manual.

Comment 5 Milan Navratil 2016-11-13 20:44:44 UTC
Hello,

I'm with the docs team and I've asked a Red Hat subject matter expert to respond to your questions.

I'll update the Red Hat Enterprise Linux 7 Storage Administration Guide based on the response from the subject matter expert.

Thank you.

Comment 6 Rob Evers 2016-11-14 21:47:26 UTC
(In reply to Milan Navratil from comment #5)
> Hello,
> 
> I'm with the docs team and I've asked a Red Hat subject matter expert to
> respond to your questions.
> 
> I'll update the Red Hat Enterprise Linux 7 Storage Administration Guide
> based on the response from the subject matter expert.
> 
> Thank you.

Due to the unpredictablity in behavior, Red Hat will still recommend against doing rescan during active IO.

We could discuss changing this to reflect that Red Hat is deferring to array vendors for such support if EMC is willing to qualify this.

Rob

Comment 7 Gopinath Marappan 2016-11-17 01:48:54 UTC
(In reply to Rob Evers from comment #6)
> (In reply to Milan Navratil from comment #5)
> > Hello,
> > 
> > I'm with the docs team and I've asked a Red Hat subject matter expert to
> > respond to your questions.
> > 
> > I'll update the Red Hat Enterprise Linux 7 Storage Administration Guide
> > based on the response from the subject matter expert.
> > 
> > Thank you.
> 
> Due to the unpredictablity in behavior, Red Hat will still recommend against
> doing rescan during active IO.
> 
> We could discuss changing this to reflect that Red Hat is deferring to array
> vendors for such support if EMC is willing to qualify this.
> 
> Rob


I found another RedHat link that suggests that scanning the bus is okay even without pausing the IOs.

https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_Linux/5/html/Online_Storage_Reconfiguration_Guide/adding_storage-device-or-path.html

In the link above, the command suggested to scan the bus is:
# echo "c t l" > /sys/class/scsi_host/hosth/scan

If this command is okay, is the following command also okay to do without pausing the IOs:
# echo "- - -" > /sys/class/scsi_host/hosth/scan

Comment 8 Rob Evers 2016-11-17 20:21:22 UTC
(In reply to Gopinath Marappan from comment #7)
> (In reply to Rob Evers from comment #6)
> > (In reply to Milan Navratil from comment #5)
> > > Hello,
> > > 
> > > I'm with the docs team and I've asked a Red Hat subject matter expert to
> > > respond to your questions.
> > > 
> > > I'll update the Red Hat Enterprise Linux 7 Storage Administration Guide
> > > based on the response from the subject matter expert.
> > > 
> > > Thank you.
> > 
> > Due to the unpredictablity in behavior, Red Hat will still recommend against
> > doing rescan during active IO.
> > 
> > We could discuss changing this to reflect that Red Hat is deferring to array
> > vendors for such support if EMC is willing to qualify this.
> > 
> > Rob
> 
> 
> I found another RedHat link that suggests that scanning the bus is okay even
> without pausing the IOs.
> 
> https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_Linux/5/
> html/Online_Storage_Reconfiguration_Guide/adding_storage-device-or-path.html
> 
> In the link above, the command suggested to scan the bus is:
> # echo "c t l" > /sys/class/scsi_host/hosth/scan
> 
> If this command is okay, is the following command also okay to do without
> pausing the IOs:
> # echo "- - -" > /sys/class/scsi_host/hosth/scan

We want to stay with what I said in comment 7.

Comment 9 Rob Evers 2016-11-28 13:38:32 UTC
Hi Gopinath,

Is EMC willing to qualify dynamic scanning under load and provide documentation indicating support?

Rob

Comment 10 Gopinath Marappan 2016-12-12 14:51:11 UTC
We are running tests along with PowerPath to qualify this scenario.  Will update as soon as they are done.

Comment 11 Marek Suchánek 2017-09-25 10:27:16 UTC
Hello Gopinath,

I'm Marek, the new Storage Administration Guide maintainer, taking over from Milan, who discussed this bug with you earlier.

Is there any update on the tests?

Thanks,

Marek

Comment 12 Gopinath Marappan 2017-09-25 11:52:44 UTC
We did some tests and did not notice any issues.  However, we cannot be sure that there wont be any issues in all scenarios, especially high IO load scenarios.

Thanks,
Gopi

Comment 13 Marek Suchánek 2017-10-17 15:57:43 UTC
Hello again Gopi,

Is it OK for me to close this bug then? Or would you like to keep it open for more research on the topic?

Thanks,

Marek

Comment 14 Gopinath Marappan 2017-10-17 16:15:07 UTC
HI Marek,
You may close the ticket.
Thanks,
Gopi

Comment 15 Marek Suchánek 2017-10-17 16:37:51 UTC
Closing this bug as INSUFFICIENT_DATA for lack of a better description.

Gopi, thanks for your help, and have a nice day,

Marek


Note You need to log in before you can comment on or make changes to this bug.