[Dxspider-support] Daily user disconnections

[ei6iz.Brendan Minish]

Ok, so I fired up SAR and set it up to capture details on the system
activity and predictably enough 0300 nothing happened. The joys of
Debugging..

It's not a given that this is a dxpsider thing, for example my virtual
server provider might be doing something on a regular schedule that
breaking open tcp sessions or causing a spike in io.

I'll carry on logging sar data to see if I have anything to go on after the
next event

On 24 April 2016 at 05:31, Michael Carper, Ph.D. <mike at wa9pie.net> wrote:

> I'm not opposed to looking into that.  But on the other hand, ALL users
> are getting disconnected... and the server is frozen when connected using
> Putty.  (In other words, I'm getting disconnected when the Putty connection
> is frozen also.)
>
> Mike, WA9PIE
>
> On Sat, Apr 23, 2016 at 9:11 PM, Lee Sawkins <ve7cc at shaw.ca> wrote:
>
>> You might not like to hear this, but HRD is the source of my frequent
>> user disconnects.   My HRD users are coming and going all the time.  I have
>> complained about this before, but nothing has ever been done.  It appears
>> if the node does not reply to HRD immediately the program will disconnect.
>> ALL other user client programs are stable and do not do this.
>>
>> Lee VE7CC
>>
>> ------------------------------
>> *From: *"Michael Carper, Ph.D." <mike at wa9pie.net>
>> *To: *"The DXSpider Support list" <dxspider-support at dxcluster.org>
>> *Sent: *Saturday, April 23, 2016 11:39:52 PM
>> *Subject: *Re: [Dxspider-support] Daily user disconnections
>>
>>
>> I don't have that line in my crontab... but I still have frequent user
>> disconnections.  I haven't yet figured it out.
>>
>> Mike, WA9PIE
>>
>> On Sat, Apr 23, 2016 at 4:47 PM, ei6iz.Brendan Minish <
>> ei6iz.brendan at gmail.com> wrote:
>>
>>> It's been bugging me for a while that ei7mre seems to kick off users
>>> every day, despite the cluster remaining running
>>> Today I did a bit of digging and this seems to coincide exactly with the
>>> cronjob that  is run by the system  from
>>> cmd/crontab at 3AM
>>> the offending entry appears to be
>>>
>>> 0 3 * * * Spot::daily()
>>>
>>> is this because the job called is taking too long and it's in the main
>>> thread? or do I have a problem elsewhere I should be investigating?
>>> Do I risk really messing things up if I change the schedule of this job
>>> to run at a time when I might be awake, to allow me to do a bit more
>>> debugging
>>>
>>> ei7mre runs on a Centos7 VM with adequate resources (20G of SSD backed
>>> Storage, 1G ram)  and the dxspider process never actually crashes (up-time
>>> is 250 days at present)
>>> ei7mre typically has ~ 140 Users at weekends these days and never feels
>>> in the least bit sluggish, very light CPU load (0.04) most of the time
>>>
>>>
>>> --
>>> 73
>>> Brendan EI6IZ
>>>
>>> _______________________________________________
>>> Dxspider-support mailing list
>>> Dxspider-support at dxcluster.org
>>> http://mailman.tobit.co.uk/mailman/listinfo/dxspider-support
>>>
>>>
>>
>> _______________________________________________
>> Dxspider-support mailing list
>> Dxspider-support at dxcluster.org
>> http://mailman.tobit.co.uk/mailman/listinfo/dxspider-support
>>
>>
>> _______________________________________________
>> Dxspider-support mailing list
>> Dxspider-support at dxcluster.org
>> http://mailman.tobit.co.uk/mailman/listinfo/dxspider-support
>>
>>
>
> _______________________________________________
> Dxspider-support mailing list
> Dxspider-support at dxcluster.org
> http://mailman.tobit.co.uk/mailman/listinfo/dxspider-support
>
>


-- 
73
Brendan EI6IZ
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.tobit.co.uk/pipermail/dxspider-support/attachments/20160424/93c1afbd/attachment.html>