C-Gate 2.8 much slower to sync networks

I have the following environment:
- 35 networks with CNIs
- C-Lution communicating via C-Gate

Previously, from starting C-Gate (whichever came with Toolkit 1.6), it took around 10-15 minutes to sync all networks so that C-Lution could take control of the lighting.

Now, having upgraded to Toolkit 1.11, it is taking *much* longer (I'm not entirely sure how long because I needed to gain control so have had to force sync using NET SYNC commands, but after 20 minutes left to its own devices I had only 5 neetworks in sync, and there were devices showing error status that were not showing errors before, including some devices which Toolkit said it was unable to communicate with).

How can I speed this up? Are the errors I'm getting an indication that there have always been errors that older versions were not detecting? (I'm refering to "state=error" in the output from a tree command. What does this actually mean?)

I am on site for the next few hours and would really like to be able to tweak some settings before I leave.

Note that there are the following differences in C-GateConfig.txt caused by the upgrade:
command.show-responses = yes (was =no)
sync-time=3600 (was 14400)

 

Sounds about right, that'll be roughly when the site was commissioned.

I'm a great believer in "if it ain't bbroke don't fix it", but at some point we're likely to be installing some newer dimmers which I'm assuming I'll need a recent TK to work with. Since I had some quiet time to do the upgrade I took a chance - ouch!

I've now rolled back to the old version but would like to persue this so I'm happy to run some more tests on my next visit.

Just so you know, all the faults have gone after rolling back. In TK1.11 I had one network I couldn't even scan (from TK, from command line, etc), but it's come straight in now I've rolled back.

As a software developer myself I know where you're coming from. I'd like to give you some decent info to work from - where do I start?

If it helps I can start incremental upgrades, it'll probably be one version per site visit (which are every couple of weeks). If so, if there are specific versions to target just let me know.

Looking at the old C-gate I have running now, almost immediately after C-gate starts I see all networks starting to synchronise (looking at the output from TREE), with all units over a period of about a minute going into NEW status (across all networks), and then over the next 10 minutes or so going to OK status as the data is collected from them.

When I did the same thing with the new C-Gate, most networks sat in new status with no units and nothing much happened until I did NET FLUSH/NET SYNC on the network manually. Even then, one network refused to do anything (sync error, no units found) whether done manually or from Toolkit.

(One problem I have is that if I scan from toolkit, all units stay in "NEW" status according to TREE, and in that state they do not seem to play well with C-Lution; I have to wait for a full SYNC to complete so all units show OK not NEW.)

As I've been investigating this I've been digging deep into the information that C-Gate can tell me. Is there any documentation about the results of the GET command? Eg for a unit which stuck in ERROR status in the new C-gate but is now OK after rolling back, I see ErrorFlags=9 - what does that mean? (The ErrorFlags value has not changed after rolling back. Another unit on the same network which showed OK status in both C-Gate versions has ErrorFlags=1 so maybe bit 5 is significant?)

 

Also:
net check_unravel
.. didn't seem to work. It's not a command I've used before, but running it before the downgrade only ever gave me "200 OK" (with nothing before that). Now, after the downgrade, I see that net check_unravel gives quite a lot of output, and it might be a useful command to know about.

I assume there's no good reason why this should have failed on my networks in the new version?

 

I'll generate some level 9 logs next time I am on site. It'll be just 1.6 and 1.11.0 for now due to the time it takes to do - I get on site around 6am and have about an hour to "play" before the site becomes too active for the lighting server to be unavailable.

Well normally TK isn't even open; the server boots, starts C-Gate automatically (with a couple of Java options to give it some extra memory) then it's left to perform the sync. About 15 mins later C-Lution starts automatically and the system should be ready to use.

I have a basic web GUI I built which I can use to monitor the progress of the syncs (basically it runs TREE commands and interprets the output) which yesterday was how I discovered that all networks were open but showing as "new" but with no units, aside from maybe 3 or 4 networks (from memory) which were starting to sync. "Normally" (ie with the old C-Gate) all would start to sync immediately and would complete within about 12 minutes. But yesterday, after waiting much longer than that I started TK to force some manual scans in order to get to a point that I could start C-Lution and get the site running (it was gone 7am by this point and that meant several scheduled lighting changes were overdue). This, combined with lots of NET SYNC commands run manually, allowed me to get operational but with one network not responding and several units showing errors.

Eventually, I had no choice but to revert back to the old version, which involved switching to a backup of the C-Gate directory and starting afresh - I was up and running within 15 mins with no unit errors.

I'll runs some clean tests next time I am on site; I'll start with level 9 logging with the new C-Gate, let it run for about 20 mins, manually sync a couple of networks after that, and shut down after about 30 mins. I'll then repeat with the old C-Gate as close as possible to the first test, but this time I'll be able to leave it running. I know that you'd probably like longer logs, but I have to get up before 4am to be there at 6am as it is! If I can get there a little earlier I will, though.

Unless you need me to run one of those I won't (at least not this time).

Not a problem, I'm very grateful that you're there to help me!

 

Thanks for this information.

What is the difference between the resulting TK scan and the auto scan? I assume that having all units showing "new" rather than "ok" means it (initially anyway) runs NET SYNCNEW - is there anything I can do to make the states change to OK?

Not sure why the C-Lution driver would care whether it's NEW or OK but it does seem to, although I haven't investigated this in much depth.

Yes, this was changed a few years back on the back of advice.

It would be good to change it back, I will do next time I am on site.

I would be very surprised if it changed any settings but yes it would be good to know.

I assume TK turns autosync back on - is that after the scan completes or on closing TK?

Since the site should run for weeks without restarting, and since there are occasions where autosync will be required (eg after a faulty unit gets replaced), it obviously matters what impact running TK has (I'm the only person who runs it, which is limited to my site visits).

That makes sense - the old config (which I reverted to) does seem to have had full logging enabled (for the past 3 years...). It's not caused a disk space issue because the log file had been set with NTFS compression, but I notice that the new version now rotates the logs which is sensible but breaks that compression (I assume I can work around this easily enough).

Obviously NTFS doesn't do as well as 7-zip (which has long been installed on almost any PC I touch, a very good tool I agree) but it still makes a big difference for log files.

 

Great, thanks! I will try when next on site (~1week)

How does this compare with the older version? I have 35 networks, presumably the other 10 will start as the first ones finish?

My objective, obviously, is to get all 35 networks in sync as quickly as possible. I appreciate that starting all 35 together is necessarily the best way to do that.

 

Hi Daniel,

The section above raises the question in my simple mind - What does Schedule Plus / Homegate do in relation to autosync= setting ?
Thanks...

 

They use autosync. The sync period is defined in the C-Gate configuration file.

 

Just realised I wrote that wrong. Obviously (I hope) what I meant was:

I appreciate that starting all 35 together is not necessarily the best way to do that.