Best practice voor rebooting HC

[Bodyart 231]

Hi ALL!

 

there were many things written about rebooting your HC.

Someone is rebooting when the RAM usage increase above certain level; some do it when the Cache memory reach certain level,m etc.

Does anyone have some guideline voor the best practice for rebooting his/her HC?

 

[Sankotronic 2,255]

Hi @Bodyart,

 

so, the best solution is that we don't need to reboot our HC's at all.

But of course due to some problems sometimes this is the only solution.

 

I have very occasionally that actors have some 1-2 second delay, but I never reboot for that reason because it occurs very rarely and I believe that is normal for system that has almost 100 included devices. What I want to say is that when this happens I don't immediately panic or get upset. I realise that wireless systems can sometimes have congestion in z-wave traffic and then this delays can happen. I write my scenes and VD's to avoid sending any extra commands if not necessary so always first checking state of the actors before sending commands. Also i always put some delays between commands if are sent to many devices at once so that devices have enough time to replay to the gateway their status.

 

I actually never  really needed to reboot my Angelina except in three cases when she get stuck and stop working. In this three cases remote or local connection was not working so I had to press Power button on her back  In all other cases I rebooted Angelina because I felt like it is OK to do that so I get rid off garbage that could collected in memory. This is always after intensive programming and testing of code. 

 

Now when I'm away from home and there is no programming and testing she is working nice. Used memory is around 44% and is fluctuating up and down for 1-2%. Buffered memory is steady on 6% and cached memory also fluctuates between 18-22% so if all used memory is included then free memory is around 30% all the time.

 

I wish I can say that Angelina is working for some time without problems, but 5 days ago we had main power outage and it lasted for several hours. Angelina did her part correctly. She send notifications when power was lost. She also turn off herself after 45 minutes to have graceful shut down. UPS that she is connected to also shut down after one hour. When power finally come back Angelina booted up properly and continue to work properly.

 

MY CONCLUSION

 

Don't reboot your HC on any small glitch that happens. If you have problems with actors that are acting with delays then rather first check your code! Sending a number of commands in z-wave network without delays and planing can cause delays. Also, I believe that turning off reporting dead devices for mains powered devices and their pooling can improve responsiveness of z-wave network, especially with big networks. That is what I did and it seems better choice that to turn on those two options. For battery operated devices pooling is useless since they will not wake up on gateway request, but to know that they are still functioning properly it is good to turn on dead device reporting.

 

So, for me rebooting is OK in this cases:

1. After extensive coding and testing of code to get rid of some garbage that can collect in the memory
2. Before firmware upgrade with same reason, to make upgrade to clean system (see @chaicka recommendation)
3. Before including more than one device to the z-wave network
4. Possibly if free memory falls under 10%, but that never happened to my Angelina (min was at 19% but still working properly)

 

[Bodyart 231]

6 minutes ago, Sankotronic said:

so, the best solution is that we don't need to reboot our HC's at all.

Hi @Sankotronic,

 

I can only agree with that!

 

Thank you very much for your extensice explanation

 

6 minutes ago, Sankotronic said:

So, for me rebooting is OK in this cases:

1. After extensive coding and testing of code to get rid of some garbage that can collect in the memory
2. Before firmware upgrade with same reason, to make upgrade to clean system (see @chaicka recommendation)
3. Before including more than one device to the z-wave network
4. Possibly if free memory falls under 10%, but that never happened to my Angelina (min was at 19% but still working properly)

 

1. Agree

2. Also agree

3. Didn't think of that and also don't quite understand why.

4. When you talk about increased memory usage do you mean cache, or all of it?

 

 

Edited June 24, 2017 by Bodyart

[Sankotronic 2,255]

33 minutes ago, Bodyart said:

3. Didn't think of that and also don't quite understand why.

4. When you talk about increased memory usage do you mean cache, or all of it?

 

3. Well, when adding bunch of devices to the network, one after another maybe it is not bad first to reboot to also clean up any garbage and cache from the z-wave network.

4. Normally only used memory should be counted for (red part of the pie), while buffered and cached memory not since they are temporary and should be released if more free memory is needed, but for some reason people often reported that their HC slow down when memory free space (green part of the pie) fell down bellow 10%, so I refer to complete usage of the memory (used, buffered and cached)

[jimicr 90]

I usually burn some candels and make a sacrifice to old and the new gods before I reboot

 

I have a reboot schedule to auto reboot at 4 in the morning on Thursdays. When there is no activity except Mablethorpe some temperature reports. 

 

I think some fw versions have memory leaks and this is my way to flush the system. 

 

Jim

Edited June 24, 2017 by jimicr

[chaicka 195]

Thanks Sankotronic for the great effort to write-up summarising the bulk of best practices which I believe most of us who are a bit more experienced with Home Center have learned over the years and been practicing (refinements after refinements based on continuous observations, analysis and knowledge sharing in this forum). It is a well written recommendation that I would place my faith on.

 

Similarly, I do not panic when slight glitches such as scene delay for a few seconds occasionally or memory utilisation increasing (very slow pace of 1-2% per few days or per week). In fact, since 4.12x onwards, I have not even reboot HC until a new (beta or official) release is out and I perform the usual 'clean-of-state' procedure (shared in forum as sticky post and mentioned by Sankotronic) prior to updating the HC software/firmware. Perhaps some day when Fibaro finally decide to adopt similar approach of automated warning and recommend reboot of Home Center if it has been up and running for xx days since last reboot, prior to software/firmware update, I will continue to practice the procedure I shared (Note: I recently purposely skipped the procedure once and after the software/firmware update, Home Center did have some weird behaviour which I still have yet to narrow down the cause - e.g. smart light scenes with 2 motion sensors tend to have a delay triggering which I have posted under '4.132 beta' thread).

 

For those who are really keen to find out the backgrounds of some of these best practices a few of us have been sharing, check out those early days of version 4.xxx alpha/beta release threads. There are a lot of 'GEMS' in those threads where memory utilisation is discussed in great length (e.g. why cache space is ok since home center is linux-based, why low free memory space does not necessarily equate to system slow/hang).