Unsolved
This post is more than 5 years old
83 Posts
2
9811
Ask the Expert: OneFS 7.1 Waikiki Preview.
Welcome to this EMC Support Community Ask the Expert conversation.
|
|
Ask the Expert: Isilon Performance Analysis Ask the Expert: AIMA: Everything you wanted to know but were afraid to ask. Ask The Expert – Isilon’s New Releases: IsilonSD Edge, OneFS.NEXT and CloudPools |
This is an opportunity to learn about and discuss the newest upcoming OneFS release code name WAIKIKI!
November brings the release of OneFS 7.1 for Isilon products, which includes a number of improvements and new features, including:
- Job Engine version 2: In OneFS 7.1, the job engine can run multiple maintenance jobs simultaneously, while dynamically modifying those jobs to ensure that impact to production workflows is minimized.
- Deduplication: OneFS 7.1 supports sharing blocks amongst multiple files via shadow stores, reducing overall space usage.
- SMB Auditing with Varonis DatAdvantage: OneFS 7.1 supports auditing of all SMB traffic to the cluster via the Common Event Enabler (CEE) framework to Varonis DatAdvantage.
- Data at Rest Encryption: OneFS 7.1 introduces support for nodes with Self-Encrypting Drives, or SEDs, allowing data to be protected against physical hard drive theft.
… along with a number of smaller features and fixes. This is your opportunity to get the skinny on the next release of OneFS!
This discussion begins on 10/7/2013 and concludes on 10/18/2013 Get ready by bookmarking this page or signing up to receive email notifications.
Your host: Jim Gossett
Work |
|
| Occupation: | Technical Support Engineer V |
| Biography: | I have a Bachelor's degree in Computer Science from the University of Washington (go Huskies!)I've worked in development, QA, IT, and support/escalations at various companies large and tiny for the past 20 years.I started at Isilon Systems as an escalations engineer in 2007, and came over to EMC when Isilon was acquired. |
| Expertise: | Currently, I'm the "Future Products Subject Matter Expert", so my general area of expertise is "what is coming next for Isilon products, and how to we train and document it?" |
| Company: | EMC |
| City: | Seattle |
| State: | WA |
| Country: | USA |
jimgossett
26 Posts
0
October 14th, 2013 18:00
Hi Bhalilov!
> Is the limitation of maximum 5 simultaneously running SyncIQ policies still valid ? If so, is this a hard of soft limit ?
This limit is still in effect. What workflow are you attempting to set up that you are looking to move past that? The limit is technically configurable, but changes to this limit can allow SyncIQ to overrun a cluster, so it is not normally supported.
> Does the new job engine have a way to control the priority if the SyncIQ Target Aware Sync ?
The Job Engine and SyncIQ subsystems are entirely separate -- there is no way to control the CPU usage of a target-aware initial sync at this time.
> Also, are there enhancement of MMC integration ?
MMC integration was not a part of the OneFS 7.1 release. The current plan is to look at tighter MMC integration sometime in 2014, but that plan is subject to change.
Sorry the answers weren't what you were looking for!
Jim
jimgossett
26 Posts
1
October 15th, 2013 09:00
Hi Peter!
> As for the SSD/GNA topic, I was referring to the setting of sysctl efs.bam.layout.ssd.sys_btree.mirror.
This is indeed wandering a bit off-topic, as it is the setting that was put into place in OneFS 7.0, but the general point is that the system B-tree is a different structure than what GNA covers. GNA covers putting copies of file/directory metadata on SSDs; as I understand it, the system b-tree covers, essentially, metadata about inodes, which are a level of abstraction above files/directories and are not covered by GNA code.
This is why there is a little bit of overhead to upgrading to 7.0 with SSDs; the usage does go up, but in general, so long as the 6.5 system was somewhere close to the supported guidelines for SSD number and density in a cluster, the amount of extra usage that mirroring system b-trees adds will not make a difference of more than a few percentage points of SSD usage overall.
JIm
pdorrington
1 Message
0
October 15th, 2013 09:00
Hi Jim,
1. If a group of folders are deduplicated and they are mirrored to another cluster using SynqIQ will the data be transferred deduped or will it get transferred and stored on the target cluster in its "fat" state ?
2. How is the Dedupe functionality licensed ? Is it per cluster, per TB or something else ?
Regards
Peter
soetingr
40 Posts
0
October 15th, 2013 13:00
Hi Jim,
For MMC, I would like to see creating a share from MMC added to just one zone. And event logging!
The network part was introduced to us on EMC World and during a one on one by David (If you want his full name mail me).
Robert
jimgossett
26 Posts
0
October 15th, 2013 13:00
Hi Robert!
> I was told some time ago that 7.1 would include a lot of enhancements on network configuration.
I'm not sure who told you those would be included in OneFS 7.1, but I'm afraid they didn't make the cut. These are still planned as improvements to the Access Zone feature (so that we can support per-Access Zone-level network configuration), but I do not have information I can share regarding when they will be available at this time.
Additional MMC integration (which I am guessing is what you are hoping for with "additional integration with standard Windows tools") also relies on similar improvements to Access Zones. Basic MMC integration works, generally speaking; what specifically are you looking for with Windows integration?
Jim
jimgossett
26 Posts
0
October 15th, 2013 14:00
Hi Peter!
> 1. If a group of folders are deduplicated and they are mirrored to another cluster using SynqIQ will the data be transferred deduped > or will it get transferred and stored on the target cluster in its "fat" state ?
In this release, SyncIQ does not have the ability to transfer deduplicated data in a deduplicated state. The data would be "rehydrated" and sent in full to the target cluster. That said, it is perfectly okay to run dedupe on the target cluster as well -- SyncIQ is smart enough to do the right thing on incremental runs even if the target has been deduplicated, and it is possible to deduplicate data that is marked read-only due to being part of a SyncIQ target.
> 2. How is the Dedupe functionality licensed ? Is it per cluster, per TB or something else ?
Per cluster, like all previous OneFS software features.
Jim
shhh1
4 Posts
1
October 15th, 2013 17:00
> On 6.5.5 it is set to 1, on 7.0 it has been found to be set to 3,
but if that is not actually the default on 7.0 or 7.1 then I don't see
an issue for us.
We have 7.0.1.1, and it was found to be set to 3. Decreased it to 1 to increase speed of SnapshotDeletes as instructed by support.
Peter_Sero
1.2K Posts
0
October 16th, 2013 02:00
Thanks Jim, and also to shhh (below)
Cheers
-- Peter
jimgossett
26 Posts
1
October 17th, 2013 11:00
Hello both Peter and shhh!
I did a bit of digging and found that there were some layout inconsistencies that were cropping up with the setting you mentioned (efs.bam.layout.ssd.sys_btree.mirror set to 3.)
As a result of these, the default setting has been reverted back to 1 (a single mirror) for that sysctl and a few others. This will solve a number of problems, not least of which is the slow performance of SnapshotDelete on clusters with fewer than 8 SSDs (this is likely the issue sssh! was running into.)
The new default is applied in 7.0.1.9, 7.0.2.4, and 7.1.0. (The prior two releases are available now.) I apologize for getting this wrong; in my defense, this *is* a pretty late-breaking change...
Jim
Peter_Sero
1.2K Posts
0
October 17th, 2013 23:00
Hello Jim,
I do appreciate your efforts as well as the findings!
Thanks a lot
-- Peter
shhh1
4 Posts
0
October 18th, 2013 05:00
Interesting!
I wonder if they are these...
efs.bam.layout.ssd.sys_btree.mirrors=1
efs.bam.layout.ssd.qab.mirrors=3
efs.bam.layout.ssd.stf_delta.mirrors=3
efs.bam.layout.ssd.logstore.mirrors=3
efs.bam.layout.ssd.sfm_delta.mirrors=3
A FlexProtect is running right now and it seems incredibly slow... on the file repair phase, and at 10 hours into it approximately only 3 million lins have been processed. Sorry for taking this off topic.
MRWA
83 Posts
0
October 18th, 2013 07:00
This Ask the Expert event has now concluded, but the conversations continue here in the Isilon Support Forum. Thanks to Jim Gossett for all his advice and answers!
Thank you to all those who participated in this event!
jimgossett
26 Posts
0
October 18th, 2013 17:00
Hi shhh!
efs.bam.layout.ssd.sys_btree.mirrors=1
efs.bam.layout.ssd.qab.mirrors=3
efs.bam.layout.ssd.stf_delta.mirrors=3
efs.bam.layout.ssd.logstore.mirrors=3
efs.bam.layout.ssd.sfm_delta.mirrors=3
Those would indeed be the sysctls -- in 7.0.1.9, 7.0.2.4, and 7.1 those will ALL be set to 1. You can feel free to make that change yourself in the meantime (although if this is a production cluster I would strongly urge you to contact Support first, just so we have a record of having suggested it.) Or you could just upgrade to the relevant maintenance release, and skip the whole "mucking about with sysctls" step.
Jim
Anonymous
5 Practitioner
5 Practitioner
•
274.2K Posts
0
October 25th, 2013 04:00
Will 7.1 support VAAI on 5.5 for NFS?
mcbrik
1 Message
0
October 27th, 2013 08:00
Hey Jim,
I just finished the 7.1 training. In regards to the Job engine, it said that 3 jobs can run. There are 2 categories, and only one job per category that can run, but there are jobs that don't fit in either category. Can you break this down for me?
What are the 2 categories and what job fall where?