(Briefly, like this question but for Windows servers.)
I have several Win2003 servers running custom application services (C/C++, not Java) that write text-based logs in a custom format.
[2009-07-17 12:34:56.7890]\t INFO\t<ThreadID>\tLog message...
[2009-07-17 12:34:56.7890]\t *WARN\t<ThreadID>\tLog message...
[2009-07-17 12:34:56.7890]\t**ERR \t<ThreadID>\tLog message...
I would like to have a way to easily and efficiently (over a not-very-fast VPN) "watch" these logs for lines that match a pattern (like tail -f |grep -E on linux). Ideally the output would need to be aggregated, not one window/shell per file or per server, and a Windows application would be best, so that I can put it in front of people who are command-line-phobic.
Any recommendations?
edit: fixed link
Try using baretail
splunk from www.splunk.com is the way to go. It is free & does exactly what you are asking for.
Related
Update: We are using AIX environment.
We have been facing some random issues with our queues (cluster queues), like:
2189 Cluster resolution error (Most frequent one)
2270 MQRC_NO_DESTINATIONS_AVAILABLE
2053 Queue full error(Weirdest) : Post one message, it will be successfully posted, post some 3-4 messages, it will throw this error
for the rest of the messages.
All these issues get resolved once we do a cluster refresh. But, I want to know the root cause, why we get these errors. What goes wrong?
How cluster refresh resolve these errors?
Could be a socket issue. You can monitor sockets according to your OS - like on windows can do
netstat -a -b -o >/newfile.txt
You could also use TCP Viewer on windows (one exe from Microsoft/ sysinternals) http://technet.microsoft.com/en-us/sysinternals/bb897437.aspx actually all the sys internal toos should be in your prod box if windows.
For sockets in linux/Un* there are other tools, some just ls commands into the RAM, depending on the version. Maybe a google will help.
Also if using windows consider moving some stuff to linux, you will have some pain in the beggining but will get better.
If this did not help you should post yor environment on your quesiton and give any other details. And if you get a jprofiler into production and use it when the issue happens.
At the very least you can do a jstack and jmap
What is version/ name of OS and of java, websphere?
If it is a socket issue can try increasing sockets (registry) and then profiling your code to see who is making too many sockets, what needs to be throttled or re-written.
Remember every page, every db connection, external cache hit (if you use) or any other URL work/ remote connection is usually a socket.
I am trying to automate the process of capturing network packets send by a particular application.I don't have problem in windows as I am using Microsoft Network Monitoring tool and that gives all the traffic send based on the process.Now the problem is that we need to achieve the same result in mac as well.
We used Wire-shark in mac for capturing network traffic after a long research but still wireshark does not capture by process name.We tried some tools that captures based on process name but with very limited information. We need the full packet informations to be captured.
What will be the best way to achieve the result in mac as like we are getting in windows?
what all parameters can be used to filter the wanted data...?
I am in fact trying to get the port number used by the process but is that a right approach ? I am not sure we can zero down based on port number.
Any help would be greatly appreciated...
Thanks in advance...
You can use lsof command, but it's more complex than netstat.
Here is a guide: lsof survival guide
I've spent hours trying to figure out the answer to this and just continue to come up empty handed. I've setup a XMPP server through OpenFire that is fully functional. My goal with creating the server was placing an alert system for when an event is completed on my server. For example when one of my renders is finished rendering (takes hours, sometimes days), it has the option of running a command when it's finished. This command would then run a .bat file telling a theoretical program to send a message via the broadcast plugin in OpenFire to all parties involved in the render. So it needs to be able to receive parameters such as %N for name of the render and %L for the label of it.
I've located two programs that do exactly what I'm looking to do but one does not work and from the sounds of the comments may have never worked and the second one is seemingly LINUX only. The render server is Windows as is the OpenFire server so naturally it would not work. Here are the links though so you can get an idea.
http://thwack.solarwinds.com/media/40/orion-npm-content/general/136769/xmpp-command-line-client/
http://manpages.ubuntu.com/manpages/jaunty/man1/sendxmpp.1.html
Basically the command I want to push is identical to that of the first link.
xmppalert.exe -m "%N is complete." %L#broadcast.myserver
This would broadcast to everyone in the labels Group that the named render is complete.
If anyone has any idea how to get either of the above links working, know of another way or simply have a better idea on how to accomplish what I'm trying to do please let me know. This is something that has been eating at me for 2 days now.
Thanks.
you can take a look at PoshXMPP which allows you to use XMPP from the Powershell.
http://poshxmpp.codeplex.com/
Alex
I have two console programs (ex. first - client, second - server).
Do Windows have a command or resource to connect it?
Client ask question, Server answer.
Anyone encountered this problem? (just win)
Client - Server is a programming model. As you are refering to programs and not scripts, this means you have to applications and you are simply have two corresponding windows instead of UI. You should look into interprocess communication, or tcp server-client, etc. in whichever lanuage you use.
//if I understood fully what you mean :)
Windows have pipes :
dir | sort
| sends data from one program to another via usual IO.
Bidirectial transfers are not that simple, unfortunately.
If you need bi-dir, you'll have to mess around sockets and stuff.
I have a farm of several physical servers each running a large number of Ruby "workers" (daemon-like processes) and I'd like to be able to monitor the health and progress of these processes from a central location, perhaps with historical graphing like Cacti provides. What's the simplest preferably-open-standard protocol for doing something like that? Please note I'm already using monit to keep the processes up and running and under control; what I'm asking for here is a single point of entry (i.e. dashboard) for checking in on them. Thanks.
If you are already using Monit then M/Monit sounds like a perfect match.
"M/Monit expand upon Monit's capabilities to provide monitoring and management of all Monit enabled hosts from one simple to use web-interface. " - http://mmonit.com/
G'day,
What about having a monitoring process on each server that checks the status of each process and then writes that out to a flat text file, say once every five minutes.
Then another process located on a central server can retrieve at those flat files and trawl through the results and flag any issues.
If you save the individual files and timestamp them, you would also be able to see any trends forming.
Just a quick ideea.
BTW The above system is used to monitor the servers in one of the largest websites in the world. Our scripts are written in Perl with a little bit of shell script but I don't see why you couldn't write your monitoring scripts in Ruby as well.
HTH
cheers,
I'd suggest to take a look at Zabbix.
It's not as simple as monit, of course, but it allows you to run data collecting agent on each of your servers, with all agents feeding the central reporting and storage server with their data. Those agents can use any custom scripts to get the metrics - you can write simple scripts to extract the data you need from your workers, send it back to the central reporting server and display it there on the dashboard.