Hi Chang, Thanks for the information, but I think my problem somewhat simpler than what you have right now. In your reply, you mention that you have at least three hosts act as ganglia server since all host A, B and C have gmond and gmetad running. Correct me if I'm wrong, gmetad is used for collecting data from all nodes that have gmond running inside them, right? But for my problem, I only have 1 server (server01) which has gmond and gmetad running inside it and all other nodes only has gmond running inside them. When I open my web browser like mozilla in server01 and display ganglia web side locally by typing: server01/ganglia, it's working. I can see all the graphs and the data looks fine to me. Then, I go to my laptop (which is still in the same LAN), I open my web browser (Internet explorer) then I typed: server01/ganglia, the ganglia web site is showing and I can browse around but most of the graphs somehow couldn't show up. Some of them can, though. This is only a hunch, is it because of the server01 was too busy collecting the data from all gmond from every node? I did try to reduce the refreshing time at gmetad.conf but it is still showing the same behaviour. Do you know what went wrong? Cause, basically, we could see the ganglia web site from any web browser from any computer inside my network, right? Thanks in advance, and I'd appreciate your response =) With kind regards, Hardian Jr. System Engineer
michael chang <[EMAIL PROTECTED]> wrote: On 9/8/05, Hardian Suprapto wrote: > Hi All, > > Before I had problem with showing graphics using web browser in localhost > cause the graphs didn't show up. Now, I can see all the graphs in ganglia > web frontend using web browser in localhost by typing: localhost/ganglia > > Now, I want to monitor it from my office using my web browser in my laptop > by typing: > server01/ganglia > lets say the host that have ganglia installed is server01. > > I can see the ganglia website but the graphs couldn't show up. Only 1 or 2 > graphs can show up. Then I go back to server01 and I go to the web browser > in server01 and as before, I can see all the graphs and they are working > just fine. > > Are there anybody who got the same experience as mine? How to overcome this > problem? You can only monitor data recieved locally -- chances are, the data is being fully transmitted one-way, or the problem you solved at one site wasn't solved at the other. I've found that even if you monitor from both locations, more often than not there is an inconsistancy; when various hosts come and go, syncronization needs to be performed (as far as I know, it isn't being performed at the moment in recent stable versions, but I haven't been able to check the very latest versions due to GCC incompatabilities). For example: Host A, B, C, all come online. All have gmond and gmetad, and A and C both have web interfaces. A and B are at location 1, C is at location 2. The three send monitor packets over a VPN interface (e.g. a tun interface provided by OpenVPN). Then, host C shuts down. Then, everyone goes down. Then host C comes up, and then B, and then A, then B goes down, and A reboots, and then C goes down, before A goes down at the end. If you examine the data, none of them will have the same data on the set. Makes for a very confusing read. That said, I'm pretty sure there's probably supposedly some scheme around this that I haven't figured out yet. What I wouldn't want to do (but may end up doing) is determining which hosts enter the network, and updating the configs manually for those -- e.g. if I have 192.168.90.1, 192.168.90.6, 192.168.90.10, and 192.168.90.14, I don't want to update the config on all machines (the four existing, plus the new one) if I add a 192.168.90.18. However, since OpenVPN seems like it's going to already require something like this (for my setup, anyways) I'll probably end up doing similarly with Ganglia. Of course, optimally, autodetection would be awesome. -- ~Mike - Just my two cents - No man is an island, and no man is unable. ------------------------------------------------------- SF.Net email is Sponsored by the Better Software Conference & EXPO September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf _______________________________________________ Ganglia-general mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/ganglia-general --------------------------------- Click here to donate to the Hurricane Katrina relief effort.
