When you mount a high availability system should always be used to avoid redundancy SPOF (Single Point Of Failure) or put another way: If you want high availability have everything you need at least 2 times because of the vital elements for the service (network card, disk, controller, fiber channel architecture, electricity ...) if you have only one, as that item is dropped availability goodbye and farewell service.
What's this? Well, come to that when we set up a high availability system before giving it to the customer we have to verify that the redundant elements that we have implemented are working properly and the service is not lost in case of failure of one of these elements. To certify the proper operation procedures involving long run cut 'the brave' of all protected SPOF. That is, having started and giving the cluster service network cables disconnected, certifies that the system is aware of the lack of that element but not affected, reconnect the cable and certify that everything is OK, and how important survive the fall is an element of a cluster as the subsequent recovery and reuse of the item once it has solved the issue. We do the same with discs, optical fiber connections ...
When these tests do I need a tool to effectively display and rapid changes of state of different elements. For example, to know the status of the link (remove and re-wire) network card I can run mii-tool
nclserver02 root @: ~ # mii-tool eth0
: Negotiate 100baseTX- FD, link ok eth1
: Negotiate 100baseTX-FD, link ok root @
nclserver02: ~ #
That tells me that at the time the 2 physical network cards are connected via cable and have link (eth1: Negotiate 100baseTX-FD, link ok
)
root @ nclserver02: ~ # mii-tool eth0
: Negotiate 100baseTX-FD, link ok eth1
: no link
root @ nclserver02: ~ #that use the application to watch, which comes standard on all distributions I've used and which frankly I find of great use, everything and so far none of the clients with whom I worked knew her. That is why I leave this post, to make known that it is this application and how to use. I'm sure that once you start using it will become one of your essential tools ;)
but I'd be better run it once and see change screen without having to re-execute the statement. This applies also to the state of the filesystems (df-k), multipath ... in short, that rather than depending on where not to go running regularly instruction that allows me to see the change.
As man prays watch, watch periodically runs an instruction screen displaying result. In the previous example I can run mii-tool
[root @ nclserver03 ~] # watch-n0, 5 mii-tool
put in reverse the changes between 2 shots
and every half second ( - n 0.5 ) watch run mii-tool and show me the result 'full screen' so that each run will overlap over the previous one, without relocation of the content. If you need to do a grep to show only relevant data I (imagine a ls-l, a cat, etc ... we quote (with the key for accents' or 'but without the' o ', not the single quote "'") to watch the sentences to run it as a whole, otherwise the instructions after a pipe (
-d
-t does not display a default header that put information on the statement being executed, the execution period and the date / time
Sources: watch - help
0 comments:
Post a Comment