XYMONSERVERS bug
list Roland Soderstrom
Hi, Xymon version 4.3.10 I have 2 identical xymon servers and specify this in the xymonclient.cfg. XYMSRV="0.0.0.0" # IP address of the Xymon server XYMSERVERS="192.168.10.1 192.168.10.2" # IP of multiple Xymon servers. XYMSRV must be "0.0.0.0". If the first mentioned server 192.168.10.1 goes down the second happily just continues and clients sends data to the second one. But, it will display the cpu on each client as yellow with "System clock is 47 seconds off (max 10)" Strange enough it is always 47 seconds on ALL clients. If I swap the ip's around XYMSERVERS="192.168.10.2 192.168.10.1" it goes green again with no time difference. I can reproduce this any time and it is always 47 seconds. Is this a bug? or expected behaviour. Cheers, - Roland
list Henrik Størner
Well, the obvious explanation would be that the clock on your two Xymon servers differ, so on the first server the time is within the clients' idea of what time it is, but on the other server it is more than 10 secs apart. Have you checked the clock on your servers ? Regards, Henrik
▸
On 27-05-2013 06:56, Roland Soderstrom wrote:Hi, Xymon version 4.3.10 I have 2 identical xymon servers and specify this in the xymonclient.cfg. XYMSRV="0.0.0.0" # IP address of the Xymon server XYMSERVERS="192.168.10.1 192.168.10.2" # IP of multiple Xymon servers. XYMSRV must be "0.0.0.0". If the first mentioned server 192.168.10.1 goes down the second happily just continues and clients sends data to the second one. But, it will display the cpu on each client as yellow with “System clock is 47 seconds off (max 10)” Strange enough it is always 47 seconds on ALL clients. If I swap the ip’s around XYMSERVERS="192.168.10.2 192.168.10.1” it goes green again with no time difference. I can reproduce this any time and it is always 47 seconds. Is this a bug? or expected behaviour. Cheers, -Roland
list Roland Soderstrom
Yes, both xymonservers sync with xntp from the same source, clients sync from the same source as well. I have manually checked time on the xymonserver and xymonclients and they are on the dot the same time. - Roland
▸
-----Original Message-----
From: Xymon [mailto:xymon-bounces at xymon.com] On Behalf Of Henrik Størner
Sent: Monday, 27 May 2013 3:11 PM
To: xymon at xymon.com
Subject: Re: [Xymon] XYMONSERVERS bug
Well, the obvious explanation would be that the clock on your two Xymon servers differ, so on the first server the time is within the clients'
idea of what time it is, but on the other server it is more than 10 secs apart.
Have you checked the clock on your servers ?
Regards,
Henrik
On 27-05-2013 06:56, Roland Soderstrom wrote:Hi, Xymon version 4.3.10 I have 2 identical xymon servers and specify this in the xymonclient.cfg. XYMSRV="0.0.0.0" # IP address of the Xymon server XYMSERVERS="192.168.10.1 192.168.10.2" # IP of multiple Xymon servers. XYMSRV must be "0.0.0.0". If the first mentioned server 192.168.10.1 goes down the second happily just continues and clients sends data to the second one. But, it will display the cpu on each client as yellow with "System clock is 47 seconds off (max 10)" Strange enough it is always 47 seconds on ALL clients. If I swap the ip's around XYMSERVERS="192.168.10.2 192.168.10.1" it goes green again with no time difference. I can reproduce this any time and it is always 47 seconds. Is this a bug? or expected behaviour. Cheers, -Roland
list Roland Soderstrom
Actually I don't have this setting in the servers at all just on the clients. So xymonserver2 doesn't know xymonserver1, more than as a client From: Ralph Mitchell [mailto:user-00a5e44c48c0@xymon.invalid] Sent: Monday, 27 May 2013 3:28 PM To: Roland Soderstrom Subject: Re: [Xymon] XYMONSERVERS bug Could that be due to waiting for the first server to time out when it's down? Ralph Mitchell
▸
On May 27, 2013 1:16 AM, "Roland Soderstrom" <user-0cec9512a49f@xymon.invalid<mailto:user-0cec9512a49f@xymon.invalid>> wrote:
Yes, both xymonservers sync with xntp from the same source, clients sync from the same source as well.
I have manually checked time on the xymonserver and xymonclients and they are on the dot the same time.
- Roland
-----Original Message-----
From: Xymon [mailto:xymon-bounces at xymon.com<mailto:xymon-bounces at xymon.com>] On Behalf Of Henrik Størner
Sent: Monday, 27 May 2013 3:11 PM
To: xymon at xymon.com<mailto:xymon at xymon.com>
Subject: Re: [Xymon] XYMONSERVERS bug
Well, the obvious explanation would be that the clock on your two Xymon servers differ, so on the first server the time is within the clients'
idea of what time it is, but on the other server it is more than 10 secs apart.
Have you checked the clock on your servers ?
Regards,
Henrik
On 27-05-2013 06<tel:XX-XX-XXXX%2006>:56, Roland Soderstrom wrote:Hi, Xymon version 4.3.10 I have 2 identical xymon servers and specify this in the xymonclient.cfg. XYMSRV="0.0.0.0" # IP address of the Xymon server XYMSERVERS="192.168.10.1 192.168.10.2" # IP of multiple Xymon servers. XYMSRV must be "0.0.0.0". If the first mentioned server 192.168.10.1 goes down the second happily just continues and clients sends data to the second one. But, it will display the cpu on each client as yellow with "System clock is 47 seconds off (max 10)" Strange enough it is always 47 seconds on ALL clients. If I swap the ip's around XYMSERVERS="192.168.10.2 192.168.10.1" it goes green again with no time difference. I can reproduce this any time and it is always 47 seconds. Is this a bug? or expected behaviour. Cheers, -Roland
list Ralph Mitchell
Yes. Your clients try to talk to server1, eventually timeout, then deliver to server2. I think the server reads the clients clock time from the status message, so if the message is delayed, the clock time looks wrong. That's a guess, I haven't looked at the source, but that would explain what you're seeing. Ralph Mitchell On May 27, 2013 1:34 AM, "Roland Soderstrom" <user-0cec9512a49f@xymon.invalid>
▸
wrote:Actually I don’t have this setting in the servers at all just on the clients. So xymonserver2 doesn’t know xymonserver1, more than as a client From: Ralph Mitchell [mailto:user-00a5e44c48c0@xymon.invalid] Sent: Monday, 27 May 2013 3:28 PM To: Roland Soderstrom Subject: Re: [Xymon] XYMONSERVERS bug Could that be due to waiting for the first server to time out when it's down? Ralph Mitchell
On May 27, 2013 1:16 AM, "Roland Soderstrom" <user-0cec9512a49f@xymon.invalid>
▸
wrote: Yes, both xymonservers sync with xntp from the same source, clients sync from the same source as well. I have manually checked time on the xymonserver and xymonclients and they are on the dot the same time. - Roland -----Original Message----- From: Xymon [mailto:xymon-bounces at xymon.com] On Behalf Of Henrik Størner Sent: Monday, 27 May 2013 3:11 PM To: xymon at xymon.com Subject: Re: [Xymon] XYMONSERVERS bug Well, the obvious explanation would be that the clock on your two Xymon servers differ, so on the first server the time is within the clients' idea of what time it is, but on the other server it is more than 10 secs apart. Have you checked the clock on your servers ? Regards, Henrik On 27-05-2013 06:56, Roland Soderstrom wrote:Hi, Xymon version 4.3.10 I have 2 identical xymon servers and specify this in the xymonclient.cfg. XYMSRV="0.0.0.0" # IP address of the Xymon server XYMSERVERS="192.168.10.1 192.168.10.2" # IP of multiple Xymon servers. XYMSRV must be "0.0.0.0". If the first mentioned server 192.168.10.1 goes down the second happily just continues and clients sends data to the second one. But, it will display the cpu on each client as yellow with "System clock is 47 seconds off (max 10)" Strange enough it is always 47 seconds on ALL clients. If I swap the ip's around XYMSERVERS="192.168.10.2 192.168.10.1" it goes green again with no time difference. I can reproduce this any time and it is always 47 seconds. Is this a bug? or expected behaviour. Cheers, -Roland
list Ralph Mitchell
I just looked at the code (xymon-4.3.10) - the xymon client program has a default timeout of 15 seconds, and retries twice. That's 45 of the 47 seconds delay you're seeing before it tries the second server. To answer your original question, yes, that would seem to be the expected behaviour... :-) Ralph Mitchell
list Roland Soderstrom
Well spotted Ralph, Thanks a lot. But the code seems to miss something here. Shouldn't it send to both servers at the same time? Regardless of the other being up or not? Is there any reason to wait for the first server to answer before sending to the next? And even so, the time from the client should be when it is sending not when the sending process starts. I really like this feature as it gives me a very cheap and easy "xymon cluster". I discovered this when one of my xymon servers died on me last week. I just had to change the http port to see the "backup server" instead. Feature request? At least we know what is happening and why. -Roland
▸
From: Ralph Mitchell [mailto:user-00a5e44c48c0@xymon.invalid]
Sent: Monday, 27 May 2013 4:49 PM
To: Roland Soderstrom
Cc: xymon at xymon.com
Subject: Re: [Xymon] XYMONSERVERS bug
I just looked at the code (xymon-4.3.10) - the xymon client program has a default timeout of 15 seconds, and retries twice. That's 45 of the 47 seconds delay you're seeing before it tries the second server.
To answer your original question, yes, that would seem to be the expected behaviour... :-)
Ralph Mitchell
list Ralph Mitchell
I don't know what it would take to run the transmit in parallel. In some
cases a response comes back from one of the servers, as mentioned in the
comments:
/*
* Even though this is the "sendtomany" routine, we need to decide
if the
* request should go to all servers, or just a single server. The
default
* is to send to all servers - but commands that trigger a response
can
* only go to a single server.
• * "schedule" is special - when scheduling an action there is no
response, but
* when it is the blank "schedule" command there will be a
response. So a
* schedule action goes to all Xymon servers, the blank "schedule"
goes to a single
* server.
*/
Perhaps the feature request should be: try each server once, then move on
to the next. Take out of the list any server that responds. Run through
the list N times as long as there are servers still in the list.
Ralph Mitchell
On Mon, May 27, 2013 at 6:12 PM, Roland Soderstrom <
▸
user-0cec9512a49f@xymon.invalid> wrote:
Well spotted Ralph, Thanks a lot.**** But the code seems to miss something here.**** Shouldn’t it send to both servers at the same time? Regardless of the other being up or not?**** Is there any reason to wait for the first server to answer before sending to the next?**** And even so, the time from the client should be when it is sending not when the sending process starts.**** ** ** I really like this feature as it gives me a very cheap and easy “xymon cluster”.**** I discovered this when one of my xymon servers died on me last week.**** I just had to change the http port to see the “backup server” instead.**** ** ** Feature request?**** ** ** At least we know what is happening and why.**** ** ** -Roland**** ** ** *From:* Ralph Mitchell [mailto:user-00a5e44c48c0@xymon.invalid] *Sent:* Monday, 27 May 2013 4:49 PM *To:* Roland Soderstrom *Cc:* xymon at xymon.com *Subject:* Re: [Xymon] XYMONSERVERS bug**** ** ** I just looked at the code (xymon-4.3.10) - the xymon client program has a default timeout of 15 seconds, and retries twice. That's 45 of the 47 seconds delay you're seeing before it tries the second server.**** ** ** To answer your original question, yes, that would seem to be the expected behaviour... :-)**** ** ** Ralph Mitchell**** ** **
list Roland Soderstrom
If the client adds the "check time" when it established contact with the server and then send data and not when the send process starts (which include any wait-time for all servers before in the list) The problem would be solved. The time we are checking is when the client sends against the time on the server. That should mean you take the time after you have established connection.
▸
From: Ralph Mitchell [mailto:user-00a5e44c48c0@xymon.invalid]
Sent: Tuesday, 28 May 2013 10:28 AM
To: Roland Soderstrom
Cc: xymon at xymon.com
Subject: Re: [Xymon] XYMONSERVERS bug
I don't know what it would take to run the transmit in parallel. In some cases a response comes back from one of the servers, as mentioned in the comments:
/*
* Even though this is the "sendtomany" routine, we need to decide if the
* request should go to all servers, or just a single server. The default
* is to send to all servers - but commands that trigger a response can
* only go to a single server.
• * "schedule" is special - when scheduling an action there is no response, but
* when it is the blank "schedule" command there will be a response. So a
* schedule action goes to all Xymon servers, the blank "schedule" goes to a single
* server.
*/
Perhaps the feature request should be: try each server once, then move on to the next. Take out of the list any server that responds. Run through the list N times as long as there are servers still in the list.
Ralph Mitchell
On Mon, May 27, 2013 at 6:12 PM, Roland Soderstrom <user-0cec9512a49f@xymon.invalid<mailto:user-0cec9512a49f@xymon.invalid>> wrote:
Well spotted Ralph, Thanks a lot.
But the code seems to miss something here.
Shouldn't it send to both servers at the same time? Regardless of the other being up or not?
Is there any reason to wait for the first server to answer before sending to the next?
And even so, the time from the client should be when it is sending not when the sending process starts.
I really like this feature as it gives me a very cheap and easy "xymon cluster".
I discovered this when one of my xymon servers died on me last week.
I just had to change the http port to see the "backup server" instead.
Feature request?
At least we know what is happening and why.
-Roland
From: Ralph Mitchell [mailto:user-00a5e44c48c0@xymon.invalid<mailto:user-00a5e44c48c0@xymon.invalid>]
Sent: Monday, 27 May 2013 4:49 PM
To: Roland Soderstrom
Cc: xymon at xymon.com<mailto:xymon at xymon.com>
Subject: Re: [Xymon] XYMONSERVERS bug
I just looked at the code (xymon-4.3.10) - the xymon client program has a default timeout of 15 seconds, and retries twice. That's 45 of the 47 seconds delay you're seeing before it tries the second server.
To answer your original question, yes, that would seem to be the expected behaviour... :-)
Ralph Mitchell