Trials and tribulations of a new Admin.

nutcase · 13 Sep 2016 at 13:42

Couldn't agree more!

We only have 17 Enterprise CALs though so most people are using them. Working on that...

nutcase · 17 Sep 2016 at 16:53

Next stage of moving the data off the original main server. 220GB at a whopping 80-140Mbps. Half a million files. Estimated 6 hours to do it.

This will make about 600GB moved off, might finally be able to defrag the thing.

edit: Forgot to use multi thread, now running at closer to 500Mbps.

nutcase · 18 Sep 2016 at 09:45

So about 750GB free, something like 42% of the drive. Defraggler still can't get it below 18% fragmented so it's now defragging the empty space.

Ugh.

nutcase · 19 Sep 2016 at 07:12

22 hours of defrag and it's only managed 4%. I hate this server

nutcase · 19 Sep 2016 at 13:48

2008 R2. 8 disk R5 array getting a massive 2MB/Sec random read off of it. It's a disaster

nutcase · 20 Sep 2016 at 13:40

P410i

It's just the built in tool in defraggler - the 100GB OS partition gives over 6MB/Sec but the data partition almost scrapes 2MB/Sec.

nutcase · 20 Sep 2016 at 14:32

It's not the overall array size that is the issue with R5, it's the individual disk size. The rule of thumb to not go above is 1TB. These are 300GB units.

And it's URE not UBER

This and the one other remainig physical server will get new disks and configured in R10 when they (eventually) get re-purposed. Which isn't yet unfortunately.

nutcase · 20 Sep 2016 at 20:24

Quartz said:
No, I meant UBER - Unrecoverable Bit Error Rate.

And with 8x 300GB drives in RAID 5 you have 2.1 TB space, so just mirror a pair of 4 TB drives. KISS. How long have you spent trying to resolve this issue? How much has that cost your employer? You can even get 4 TB SSDs these days; the Samsung SSDs are around £1200, so for £2500 you could buy a pair, resolve the issue, and gain a massive performance increase.

So Non-recoverable read errors per bits read? Not seen it referred to as UBER before. But same as hitting an URE.

Actual time spent on it? A couple of hours at the very most. Time the software has spent running? Couple of hundred hours probably. Amazed a disk hasn't coughed to be honest. It's 3 steps forward and 1 step back as I'm getting it down to a couple of hundred fragmented files then over the course of the working day (~05:00-~22:00) it goes back up to tens of thousands. Being that it was over half a million originally progress has been made.

Think I'd get rather bizarre looks if I tried to get the OK to spend ~£2100 on consumer drives for a server that don't have the right firmware for the server's inbuilt management.

nutcase · 20 Sep 2016 at 21:03

Caged said:
Why are you having to work with a complete sack of crap for your hardware? Surely it's a better use of your time and the time of everybody that works at your company to invest in new stuff that's more suitable for the workload?

5 years ago the servers were all dumped in a pile on a desk for the world to see. It's taken a lot of persuasion to get a room built, a rack, aircon and so on. The company was a very small one that expanded fast. IT was a necessary evil that they hadn't realised was actually central to its operation. This mentality is changing, and now I've gone from occasional poking to pretty much in charge of IT it's changing fast. But in the last 24 months, we've spent more on IT than any other part of the company has spent. I've had to be gentle with expenditure

I've personally spent about £25k - split between new PC to replace XP ones, a whole new backup regime and a new host. We got audited by microsoft and with misunderstanding how things work (not by me, not that I understand their licencing!) we had quite a large bill - number of activations does not mean you can use it that many times at once :rolleyes:

When I took it on I said if you give me £100k I'd built a whole new system that does what they need to do. Give me less and it will be a slow process with hiccups on the way. That's what they're getting. I'm enjoying the challenge, but some things do my head in.

nutcase · 20 Sep 2016 at 21:06

Ricimer said:
A few things, random read/write on any traditional disk array will be that slow, sequential throughput could be fine. You can easily get 1-2MB/s 100% random but still several hundred MB/s sequential r/w.

Defraggler is pretty terrible from what I've seen of it. Turn on scheduled defrag in windows for once a week and worry less about the stats, files will fragment.

Absolutely. But when a different partition on the same array can churn out about 3 times the data you know things can be improved - for no cost.

I've had no issue with defraggler.

nutcase · 21 Sep 2016 at 08:02

nutcase_1uk said:
One more item to add to the pile (and a very big pile it is) is our P-WAN renewal is coming up. Another company that supplies our IP Phones has said they can reduce our costs significantly (~30%) with no install costs. We currently have 3 sites with fibre, and 4 with EFM.

Turns out they're offering FTTC for all but a couple of sites. When the current setup was being planned, FTTC wasn't considered because because there was no SLA, no guarantees on bandwidth contention or priority.

The telephone supplier are saying they can offer 22hour fix (with backup ADSL as we have currently), and QoS for our traffic as well as traffic type QoS. Our current supplier still only offers FTTC as one step up on a domestic product. Currently we have 4 or 9 hour fix for our primary circuits.

Two sites have the cabinet within spitting distance so we are offered the full 80/20 rate. Which is WAY over our needs (head office has 20mbit and rarely hits it)

Has anyone looked at FTTC as an alternative to traditional leased lines? Are they a viable alternative now? I'm very sceptical as what we have now works very well indeed but that kind of saving is not to be sniffed at. The supplier before the one we have now we had awful latencies etc. and I *really* don't want to go back to those days!

Just found out another reason for them being cheaper - they have only included a backup connection for the main office, not for any of the satellites :rolleyes:

so the above mention of FTTC with backup ADSL was wrong.

nutcase · 21 Sep 2016 at 08:52

Ricimer said:
Defraggler likes to move *everything* each time I run it, while the built in tools will leave slightly more files fragmented, they are more conservative during consolidation so usually take much less time.

There will be other things to consider, file locks, antivirus, physical/virtual etc (things like many dynamic vhds expanding by small amounts on the same dive causing fragmentation on the HV side).

I've always found the performance monitor with queue length the best indicator for issues in windows, along with iomonitor being good to check new builds.

Although I desperately miss esxtop when using hyper-v servers :|

Oh I know it will never get down to nothing, and that's fine. The build in defrag tool wasn't touching it at all.

Percent disk time is never below 100, and often up in the high hundreds/low thousands. Queue length for the most part isn't too bad (that's where it being an 8 disk array helps), normally between 1 and 10.

nutcase · 21 Sep 2016 at 19:15

nutcase_1uk said:
Just found out another reason for them being cheaper - they have only included a backup connection for the main office, not for any of the satellites so the above mention of FTTC with backup ADSL was wrong.

They've got the hint now and are re-speccing so we're not downgrading connections.

Meantime had a phone call today from a different company - "Virgin Media are upgrading firmware on some of their tails are they can't connect to yours". Er that will be because we've not been a client of yours since late 2013. Seriously, how do some companies survive if they hadn't noticed a connection been disconnected for nearly 3 years? Makes me look competent!!

nutcase · 14 Nov 2016 at 14:25

Things are a little quiet at the moment because I'm just getting on with it, but mainly because I'm not allowed to spend any money at the moment :mad:

Have had a couple of emails sent to the directors recently singing my praises. One was I suspect trying to placate me as I explained the same thing many times in really easy language and they still didn't get it - "switch user"... Then it finally clicked.

The other one I guess I actually did something right?

Having a bit of trouble though - there's a decorator in doing a couple of rooms and so far they've ripped one network socket out the wall, and connected one end of a patch lead into a wall port, and the other end into another wall port. Fortunately I recently put in protection against that.