Network connection drops after non-use

[JohnFLi]

I am still working on setup, but one thing I have noticed.....
I have a cifs share, and I have a pc that has the share open. Not transfer files or anything, just windows explorer open to the share.
After x amount of time (not sure on the exact amount) when I go to click on that share, there is a delay before anythign happens. Almost like the share had disconnected with non-use. I have had this with multiple pc's.

Any thoughts?

FreeNAS-9.10.1 (d989edd)
PlatformIntel(R) Xeon(R) CPU E5-2620 v4 @ 2.10GHz
Memory65393MB
System TimeWed Oct 19 09:00:48 PDT 2016Uptime9:00AM up 18 hrs, 0 usersLoad Average0.54, 0.16, 0.05

 

[pirateghost]

Do you have your disks set to spin down?

 

[JohnFLi]

ummmmmm, not that I know of.

 

[pirateghost]

You have 24 - 10 TB disks on 64gb RAM?

 

[JohnFLi]

yup currently only using 18 drives (3 vDevs of 6)

 

[Noctris]

Have you done a full disk check of each disk ( smart) ? I had a server with a couple of different pools where 2 of them where brand new disks and 1 was an old pool from a different machine to transfer the data. 2 of the disks in the old pool were a bit wonky ( a bunch of reallocated sectors). Even though freenas gave it as healthy, i experienced similar "delays" as you do. After transferring the data and removing the disks from the system, performance increased dramatically.

So i chalked it up to 2 possible things:

A) it was because my server was a little light on ram for the 3 pools ( although 32GB of ram for a 2X3TB mirror, 7 x 3TB RaidZ2 and the old 7 X 1 TB RaidZ2 looks like enough to me)

or

B) It was because of the lesser then great disks which caused a delay because of the problem.

No certainty , just an experience.

 

[Stux]

Have you done a full disk check of each disk ( smart) ? I had a server with a couple of different pools where 2 of them where brand new disks and 1 was an old pool from a different machine to transfer the data. 2 of the disks in the old pool were a bit wonky ( a bunch of reallocated sectors). Even though freenas gave it as healthy, i experienced similar "delays" as you do. After transferring the data and removing the disks from the system, performance increased dramatically.

So i chalked it up to 2 possible things:

A) it was because my server was a little light on ram for the 3 pools ( although 32GB of ram for a 2X3TB mirror, 7 x 3TB RaidZ2 and the old 7 X 1 TB RaidZ2 looks like enough to me)

or

B) It was because of the lesser then great disks which caused a delay because of the problem.

No certainty , just an experience.

The solnet array test script is good for finding these performance issues. It's part of the disk burnin procedure

 

[Noctris]

The solnet array test script is good for finding these performance issues. It's part of the disk burnin procedure

I know, but since it was an old pool that just needed to be copied once onto a new machine, i didn't really bother. I knew there were disks nearing the end of lifetime and decided to build a new machine when the first disk was starting to realloc sectors. When setting up the old machine i tested it with the solnet script ( and i did with the new one too).

But for the new build of OP, i would definitely do a burn in test. I've seen many, many, many cases where , despite being new, the disks were flawed due to manufacturing or just rough delivery.

Had several bad experiences like raid-5 sets never completing their initial build with brand new disks or worse, failing right after being taken in production.

 

[JohnFLi]

ok, so how do I go about 'testing' the disks?

 

[Stux]

It takes a few days to run the burnin procedure. But it can be done across all disks at once.

https://forums.freenas.org/index.php?threads/how-to-hard-drive-burn-in-testing.21451/

Note: badblocks is destructive by default!

 

[JohnFLi]

Ok..... the unit is not in production yet so no worries