Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > comp.lang.clipper.visual-objects > #9099 > unrolled thread

Temporay network dropouts and ADS

Started byMartin <Martin@spam.spam>
First post2016-03-07 10:50 +0000
Last post2016-03-08 13:56 -0700
Articles 9 — 5 participants

Back to article view | Back to comp.lang.clipper.visual-objects


Contents

  Temporay network dropouts and ADS Martin <Martin@spam.spam> - 2016-03-07 10:50 +0000
    Re: Temporay network dropouts and ADS dlzc <dlzc1@cox.net> - 2016-03-07 06:12 -0800
      Re: Temporay network dropouts and ADS Martin <Martin@spam.spam> - 2016-03-07 14:20 +0000
        Re: Temporay network dropouts and ADS dlzc <dlzc1@cox.net> - 2016-03-07 07:54 -0800
          Re: Temporay network dropouts and ADS Martin <Martin@spam.spam> - 2016-03-08 09:31 +0000
    Re: Temporay network dropouts and ADS "Michael Hagl" <haglmi@hagl.de> - 2016-03-08 08:01 +0100
      Re: Temporay network dropouts and ADS "Joachim Duerr" <jojo.duerr@gmx.de> - 2016-03-08 17:14 +0000
        Re: Temporay network dropouts and ADS "Michael Hagl" <haglmi@hagl.de> - 2016-03-09 08:28 +0100
    Re: Temporay network dropouts and ADS Oskar Atkinson <Oskar@atk-it.com> - 2016-03-08 13:56 -0700

#9099 — Temporay network dropouts and ADS

FromMartin <Martin@spam.spam>
Date2016-03-07 10:50 +0000
SubjectTemporay network dropouts and ADS
Message-ID<nbjm8j$7eo$1@dont-email.me>
Hello All

I have no idea why but now we seem to get a lot more network issues at 
customer than we used to quite a few years ago.

We get servers going unresponsive for 5 or so seconds (this was a backup 
but other SW survivied), we are getting network dropouts.

The problem we have is getting VO programs to behave and properly 
reconnect after the incident is over. The other applications such as 
email and telephone reestablish, but VO apps either stay in a zombie 
state or just crash.

Does anyone have any tips please to make a VO/ADS application more 
resilient?

Thanks

Martin

[toc] | [next] | [standalone]


#9100

Fromdlzc <dlzc1@cox.net>
Date2016-03-07 06:12 -0800
Message-ID<f2f2a257-bb17-4f60-ada4-9b58a16e6342@googlegroups.com>
In reply to#9099
Dear Martin:

On Monday, March 7, 2016 at 3:50:52 AM UTC-7, Martin wrote:
...
> I have no idea why but now we seem to get a lot more
> network issues at customer than we used to quite a
> few years ago.
> 
> We get servers going unresponsive for 5 or so seconds

What server OS today, and what server OS a "few years ago"?

> (this was a backup but other SW survivied), we are
> getting network dropouts.

Sounds like there is excessive traffic (virus), failing switches / routers, or failing cabling.

> The problem we have is getting VO programs to behave
> and properly reconnect after the incident is over.

You are not using WiFi here, are you?

> The other applications such as email and telephone
> reestablish, but VO apps either stay in a zombie 
> state or just crash.

Those applications start a session, then essentially close all files at the end of a transaction.  If you code that way also, so that all non-local DBFs (even drive mappings) are closed, and established only when being used, you will be less likely to zombie.  And of course, what do your error handlers look like?

> Does anyone have any tips please to make a VO/ADS
> application more resilient?

I don't do VO...

David A. Smith

[toc] | [prev] | [next] | [standalone]


#9101

FromMartin <Martin@spam.spam>
Date2016-03-07 14:20 +0000
Message-ID<nbk2hs$ni$1@dont-email.me>
In reply to#9100
OSes

Well a few years ago they were mainly Netware, now everyone is using 
newer versions of Windows.

There is a known dodgy switch at one site, but it tends to clear after a 
few seconds and it is only our software playing up.

As to open close, that would increase the running times and program 
complexity, the routines are monitoring barcode kit and are failing when 
ADS is available again.

No errors going into error logs unfortunately

Definately no WiFi!

On 07/03/2016 14:12, dlzc wrote:
> Dear Martin:
>
> On Monday, March 7, 2016 at 3:50:52 AM UTC-7, Martin wrote:
> ...
>> I have no idea why but now we seem to get a lot more
>> network issues at customer than we used to quite a
>> few years ago.
>>
>> We get servers going unresponsive for 5 or so seconds
>
> What server OS today, and what server OS a "few years ago"?
>
>> (this was a backup but other SW survivied), we are
>> getting network dropouts.
>
> Sounds like there is excessive traffic (virus), failing switches / routers, or failing cabling.
>
>> The problem we have is getting VO programs to behave
>> and properly reconnect after the incident is over.
>
> You are not using WiFi here, are you?
>
>> The other applications such as email and telephone
>> reestablish, but VO apps either stay in a zombie
>> state or just crash.
>
> Those applications start a session, then essentially close all files at the end of a transaction.  If you code that way also, so that all non-local DBFs (even drive mappings) are closed, and established only when being used, you will be less likely to zombie.  And of course, what do your error handlers look like?
>
>> Does anyone have any tips please to make a VO/ADS
>> application more resilient?
>
> I don't do VO...
>
> David A. Smith
>

[toc] | [prev] | [next] | [standalone]


#9102

Fromdlzc <dlzc1@cox.net>
Date2016-03-07 07:54 -0800
Message-ID<46421bd2-95a1-4598-9518-5a54606ef3b6@googlegroups.com>
In reply to#9101
Dear Martin:

On Monday, March 7, 2016 at 7:20:38 AM UTC-7, Martin wrote:
...
> Well a few years ago they were mainly Netware,
> now everyone is using newer versions of Windows.

There is a problem you can do little about.  A CNE might be able to optimize services for the affected servers.  What other applications are they serving, besides ADS?  I wonder if ADS could be implemented on a standalone Linux box?

> As to open close, that would increase the running
> times and program complexity, the routines are
> monitoring barcode kit and are failing when 
> ADS is available again.

OK, but the applications you cite, are coded to do that very thing.

Create a routine that records all files open, all selected indexes, all relations between files, and current record in each.  Add a timeout, and call this routine (a PUSH) to record state, and close all those remote files.

Then on wakeup, do a POP, reestablish all those files and positions.  It might take a few seconds to run, but so do email and phone apps.

Do the programs go "zombie" after 1 minute, or after 30 minutes or more without use?

Have you looked for memory leaks in your application as it runs?

I cannot solve this, I am just asking questions to get your brain running.

How about running your code *on* the server, and use Terminal Services?  These are usually much more hardy for flaky networks / servers.

David A. Smith

[toc] | [prev] | [next] | [standalone]


#9104

FromMartin <Martin@spam.spam>
Date2016-03-08 09:31 +0000
Message-ID<nbm5v2$tu3$1@dont-email.me>
In reply to#9102
Thanks David

Well the "Zombie" state is random, when the network is behaving they can 
be running undisturbed for weeks on end, I did find one condition and 
have fixed it by shutting down and immediately restarting the executable 
concerned. But I had the other front end sitting there in a funny state 
for two hours, until I woke it up, it crashed, and auto restarted, then 
sent two hours of barcode scans to the server engine.

front ends and back ends use IP for wake up messages, and pass data via 
a shared DBF.

I was waking them up by asking them to disconnect (close the dbfs) then 
reconnect (reopen them) which is used when we database update our software.

On reconnect there is a routine to send all the stored barcodes back to 
the server.

I will need to put more thinking onto this, why they are going into this 
mode.

Martin

On 07/03/2016 15:54, dlzc wrote:
> Dear Martin:
>
> On Monday, March 7, 2016 at 7:20:38 AM UTC-7, Martin wrote:
> ...
>> Well a few years ago they were mainly Netware,
>> now everyone is using newer versions of Windows.
>
> There is a problem you can do little about.  A CNE might be able to optimize services for the affected servers.  What other applications are they serving, besides ADS?  I wonder if ADS could be implemented on a standalone Linux box?
>
>> As to open close, that would increase the running
>> times and program complexity, the routines are
>> monitoring barcode kit and are failing when
>> ADS is available again.
>
> OK, but the applications you cite, are coded to do that very thing.
>
> Create a routine that records all files open, all selected indexes, all relations between files, and current record in each.  Add a timeout, and call this routine (a PUSH) to record state, and close all those remote files.
>
> Then on wakeup, do a POP, reestablish all those files and positions.  It might take a few seconds to run, but so do email and phone apps.
>
> Do the programs go "zombie" after 1 minute, or after 30 minutes or more without use?
>
> Have you looked for memory leaks in your application as it runs?
>
> I cannot solve this, I am just asking questions to get your brain running.
>
> How about running your code *on* the server, and use Terminal Services?  These are usually much more hardy for flaky networks / servers.
>
> David A. Smith
>

[toc] | [prev] | [next] | [standalone]


#9103

From"Michael Hagl" <haglmi@hagl.de>
Date2016-03-08 08:01 +0100
Message-ID<nbltc1$obc$1@news.albasani.net>
In reply to#9099
Hi Martin,

have you tried to play with entries in ads.ini?

sample:
***************************************************
[SRV2011]
LAN_IP=192.168.115.2
LAN_PORT=6262

[SETTINGS]
RETRY_ADS_CONNECTS=2
MAX_TIMEOUTS=40
***************************************************
On virtualized 2012-Server, ADS-Init had sometimes a problem to find server
The settings are described in ads help file.

Michael Hagl




"Martin" <Martin@spam.spam> schrieb im Newsbeitrag 
news:nbjm8j$7eo$1@dont-email.me...
> Hello All
>
> I have no idea why but now we seem to get a lot more network issues at 
> customer than we used to quite a few years ago.
>
> We get servers going unresponsive for 5 or so seconds (this was a backup 
> but other SW survivied), we are getting network dropouts.
>
> The problem we have is getting VO programs to behave and properly 
> reconnect after the incident is over. The other applications such as email 
> and telephone reestablish, but VO apps either stay in a zombie state or 
> just crash.
>
> Does anyone have any tips please to make a VO/ADS application more 
> resilient?
>
> Thanks
>
> Martin 

[toc] | [prev] | [next] | [standalone]


#9108

From"Joachim Duerr" <jojo.duerr@gmx.de>
Date2016-03-08 17:14 +0000
Message-ID<dk8fj7F4mtvU1@mid.individual.net>
In reply to#9103
Michael Hagl wrote:

>RETRY_ADS_CONNECTS=2

only 1 and 0 are defined values for that setting.

-- 
Joachim Duerr Softwareengineering
http://www.jd-engineering.de

[toc] | [prev] | [next] | [standalone]


#9121

From"Michael Hagl" <haglmi@hagl.de>
Date2016-03-09 08:28 +0100
Message-ID<nbojb9$rpe$1@news.albasani.net>
In reply to#9108
Hi Joachim,

oh sorry, your are right.
2 is nonsense

Michael Hagl


"Joachim Duerr" <jojo.duerr@gmx.de> schrieb im Newsbeitrag 
news:dk8fj7F4mtvU1@mid.individual.net...
> Michael Hagl wrote:
>
>>RETRY_ADS_CONNECTS=2
>
> only 1 and 0 are defined values for that setting.
>
> -- 
> Joachim Duerr Softwareengineering
> http://www.jd-engineering.de 

[toc] | [prev] | [next] | [standalone]


#9116

FromOskar Atkinson <Oskar@atk-it.com>
Date2016-03-08 13:56 -0700
Message-ID<n2fudb1q20mc61elhn3uauf7o1hdqcfr78@4ax.com>
In reply to#9099
Martin, did you see this:

>>
Checklist for OS related performance issues with ADS
Check the power plan that in use on Windows Server 2008 R2. The default power plan with a clean install of this OS is the Balanced plan. Using this plan (or other power-savings plans) can have a
significant effect on Advantage Database Server performance. The negative performance impact seems to be even bigger when CPU utilization is low.

If you think this is the case:
1.        Change the Power Option to "High Performance" in Windows. Go to the Control Panel, select Power Options, and then choose the "High performance" plan. Alternatively, I think it might be
sufficient to change only the Minimum processor state setting, but I have not tested that extensively.
2.        Also check the BIOS settings and look for something called "CPU Power Management" or possibly "Demand-Based Power Management". If it is enabled, then disable it. This setting also in some
cases compounds the problem and seems to be somewhat independent of the OS power management settings.
3.        If running under VMWare, check the Power Management option (under the Configuration tab for the Host server). Make sure the policy is set at "High Performance". 
Performance will, of course, vary in many different situations, but testing on one specific machine showed that the above changes (1 and 2) caused a test that appended records to run much faster. With
all power management settings enabled, the test took 33 seconds to append 100,000 records. With the power management turned off, the same test ran in less than 7 seconds.
More details about this are in a Stackoverflow question.
<<

Oskar



On Mon, 7 Mar 2016 10:50:53 +0000, Martin <Martin@spam.spam> wrote:

>Hello All
>
>I have no idea why but now we seem to get a lot more network issues at 
>customer than we used to quite a few years ago.
>
>We get servers going unresponsive for 5 or so seconds (this was a backup 
>but other SW survivied), we are getting network dropouts.
>
>The problem we have is getting VO programs to behave and properly 
>reconnect after the incident is over. The other applications such as 
>email and telephone reestablish, but VO apps either stay in a zombie 
>state or just crash.
>
>Does anyone have any tips please to make a VO/ADS application more 
>resilient?
>
>Thanks
>
>Martin

[toc] | [prev] | [standalone]


Back to top | Article view | comp.lang.clipper.visual-objects


csiph-web