Mom does not read or speak mandarin, of course no english. Worry about her trip from Shanghai. If anybody fly from Shanghai to New York or Newark, please contact me by mail. thanks.
Don't try to reinvent wheel. Search Flume, SumoLogic and Splunk. Pick one with your budget. The difficult part is not logging infrastructure. Global tracing is not trivial. Once you have server applications with lots of async operations, instrumentation and log analytics are not easy.
writing against Cassandra is not difficult, maintaining a Cassandra cluster is a different story. All I can say is Cassandra is better for this job, particularly on ec2.
That's the part that a custom api may be easier for this job. You can log stuff with a list of tags and index them. log doesn't have to be unstructured.
【在 c****e 的大作中提到】 : Don't try to reinvent wheel. Search Flume, SumoLogic and Splunk. Pick one : with your budget. : The difficult part is not logging infrastructure. Global tracing is not : trivial. Once you have server applications with lots of async operations, : instrumentation and log analytics are not easy.
w*z
46 楼
second this. If you have different languages writing to Cassandra, the dev cost is pretty high. Are you going to implement the API's by yourself? Ruby, Php, Java? How do you maintain it going forward? If you really want to do it yourself, it's better to provide rest service to encapsulate the access layer to C*. It's not going to be fun if you have different clients writing to C*. What if you want to change your schema, you need to change all the client code? And C* doesn't have good ui too to query the data, it's has CQL and CLI, both are command line tool.
【在 c****e 的大作中提到】 : Don't try to reinvent wheel. Search Flume, SumoLogic and Splunk. Pick one : with your budget. : The difficult part is not logging infrastructure. Global tracing is not : trivial. Once you have server applications with lots of async operations, : instrumentation and log analytics are not easy.
d*r
47 楼
是的,我打算用一种通用格式的API(e.g. REST)包装 log server 的 service,然后 AWS内网的其他 hosts,call 这个API就是了。
pretty to
【在 w**z 的大作中提到】 : second this. : If you have different languages writing to Cassandra, the dev cost is pretty : high. Are you going to implement the API's by yourself? Ruby, Php, Java? : How do you maintain it going forward? : If you really want to do it yourself, it's better to provide rest service to : encapsulate the access layer to C*. It's not going to be fun if you have : different clients writing to C*. What if you want to change your schema, : you need to change all the client code? : And C* doesn't have good ui too to query the data, it's has CQL and CLI, : both are command line tool.
For logging, you can use simple structure like this. log CF TimeUUID as key, your log stuff tag CF tagName:tagValue as key, all log keys for this tag So you can search by tagName:tagValue + time range. Now if you only want a time range search. Put all log keys in an index CF. Range search is available on row level in column names. So you just need to come up with your own index.
【在 o*******0 的大作中提到】 : Cassandra 的 index 只能做点查询,支持不了范围。你怎么解决这个呢?