VDP Now uses constant bandwidth even during blackout period

[LoneCrowe]

The initial issues I've had with VDP forced me to start fresh and make a completely new guest.  I deployed an ova for the .5TB VM.  I installed it to an NFS datastore.

It all installed fine and is working.  I can connect.  I have not done a thing.  I didn't run an integrity check, or set up  backup or anything.  My graphs for bandwidth usage to the NFS datastore shows there is constant traffic going there even though it is in its blackout period.

I had a similar experience the first time I installed it.  It would use 100Mb of transfer on the switch to the datastore for doing NOTHING.   Now I used a small older datastore its only using about 40Mbit but still using traffic.

Is this intentional?  I didn't see anything about this in the documentation.

The tcpdump shows that VDP is contacting the datastore.  The datastore when I tcpdump there it shows coming from the HOST ip (not the vdp ip)  which is curious.

Anyone explain to me why this would be needing to touch the datastore when nothing is needed to be done?

[LoneCrowe]

Here is an example you can see pre - installation to a data store and post.. this amount of traffic will go on for at least 24 hours like it did to the other datastore.

Why?

[LoneCrowe]

Think I should put in a support ticket for this?  Or is anyone else having the same issue?  Or is it just me?

[Paul11]

The VDP-appliance has a very high "read rate" with up to 40.000 KBps every 5 minutes for about a minute (on an iSCSI Storage) .

Than there is no traffic for the next 4 minutes, and this goes the whole day.

Don't know what they are doing every 5 minutes with so much data.

Paul

[LoneCrowe]

Well there should be something in the documentation for this.  Thanks for the answer though.  If it needs to continually read the hard drive / data store I might as well put VDP on a machine that has a local hard drive instead of eating up bandwidth on the switch...

But this sort of defeats the purpose of having any HA..

[SCX]

I am seeing the same "spikes" in bandwidth for 5 minutes at a time, every hour or so. It'd be nice to get an answer as to what exactly is going on here. And yes, putting vDP on a local drive does kind of defeat the point of having backups in the first place.

[GSparks]

The "generic" answer to this is that we have constant communication between the VDP appliance and vCenter.

There are some levels of communication that occur constantly - such as the creation and monitoring of events as well as inventory queries that occur as a result of vCenter events.  There are also queries that occur on a recurring basis - such as polling vCenter for known hosts, and independent queries that are run to validate the status of VDP with vCenter.

[LoneCrowe]

Well then,  I guess VDP is out as an option for me.  Unless I put it on a machine with local storage.  I can't have it eating up Mbits of bandwidth even though its a Gigabit switch, its additional overhead that is unecessary in smaller networks.

[SCX]

Communication between devices is fine and expected. What isn't fine is the profligate amount of bandwidth that this communication consumes.

[mobcdi]

Do these spikes have any relation to the number of hosts or vm's in your environment or the rate of change in your inventory or is it pretty much constant for any setup?

[LoneCrowe]

Like I said I JUST INSTALL VDP to a remote host and not set up any backups or anything.  I connect it to Vcenter obviously which has maybe 20 guests between 3 servers, and the bandwidth starts to go.  Depending on the size of the install, the more bandwidth it does.

For example I installed the 1TB version and it was doing about 100Mbit of data on a gigabit switch non stop.   Then I moved it to a smaller freenas box and installed the 500GB version and now it does about 30Mbit non stop.

[DanielOlofsson]

Interesting, I was just lurking the community to find something about this because I have noticed the same "issue".

I deployed VDP yesterday and my graphs shows a constant ~19 MB/s transfer.

See attached screenshot from inside the vSphere Client (the peak at 21:00 is when a test backup started of 4 machines)

My rrd-graphs shows same measures on the ports used for VDP.
My VDP is deployed via iSCSI 1 Gbit/s, I used the 2TB ova.

Edit:
Added a screenshot from the Web Client monitoring real-time disk usage which is showing short bursts at around 70 MB/s
No other VM in my cluster is doing this.

Added another screenshot from the vSphere Web Client

[Buschfunker]

Any use relating this story?

Same issue here. 2TB VDP and 200-400Mbit Peaks every 5 Minutes.

Why?

[calypsocraig625]

Exactly the same issue here - anyone anyone?

[bmannz]

We have exactly the same issue.  2TB VDP appliance on an NFS volume.  It peaks at 300Mbit every 5 minutes.

We are considering alternate options and will most likely move away from VDP.

[Buschfunker]

As we can see VMWare does not react on any of this posts here.

Why they don't say "Sorry, the VDP is a bad product. We'll not sale this anymore. Any customer who has bought VDP should use another product".

But the don't say anything to this. No Updates - Nothing.

What are the alternatives?

[ajbase]

Same problem here.  2TB VDP.  Seeing the 200-400Mbit spikes.  Averaged out over the day the utilization is 175 Mbps.  The admin guide claims "Minimizes network traffic by deduplicating and compressing data," but it doesn't seem to be working as advertised.  I wouldn't call 175 Mbps all day long "minimal."

[djet]

A support rep told me VDP runs storage performance test every 5 minutes.

[matteowiz]

Is there a way to disable it or to tune? 5 minute is very strange imho

regards

--

matteo

[Reto_Kne]

Same Problem... disable perfbeat is not supported, but you can change it to 60mins.

Answer from support:

The developers have come back to me regarding disabling perfbeats.  We are not allowed to give out unsupported configurations.  If you were to do this then you would have to this yourself.  To change the perfbeat timeouts it can be set to a maximum of 1 hour.  By default it is set to 5 minutes (300 secs.)

The command to change the value to an hour is

#avmaint config perfinterval=3600 --ava

From a support point of view there anything further you need at present?  It has been a pleasure dealing with you on this case and if it is ok with you we can achieve the case for now and reopen or reference it in the future should the need arise.

regards

Reto