Open matteoredaelli opened 13 years ago
Yeah the HTTP interface currently doesn't do everything, such as adding or deleting data. There is already another feature request about being able to add data over HTTP (#35), being able to create metrics over HTTP (#79) although I'm not sure you'd need this, and being able to delete data over HTTP (#81, which you filed).
Other than the above, is there anything else you'd need? Is there any reason why you'd want an HTTP interface or would you also be happy if it was another TCP-based protocol?
Hi
An HTTP interface would be more easy and standard.. many programming languages have an http library (usually curl based). if not possibile, a well documented pure TCP interface would be usefull..
But also a TCP interface would be nice, if you prefer it or take less time to implement it..
Regards Matteo
Hi, I'd like to reopen this and ask for advice from anyone who's thought about the best way to pull from OpenTSDB into R. I'm familiar with R package development, and with C/C++ (some) but not with the layers under OpenTSDB, like Hadoop and Thrift.
Is it worth building on top of RHadoop or is there a level of abstraction between Hadoop and the HTTP/JSON interfaces others have built that might be a sweet spot for a read-only interface?
I've put together a (read-only) interface as an R package over at https://github.com/holstius/opentsdb. Seeking comment.
It's not optimized, only uses HTTP, but maybe could serve as a reference implementation (or straw man) for an R/OpenTSDB interface, whether read-only or more fully featured. If anyone would like to fork or contribute, please let me know!
That seems fine to me. Maybe you'll get more feedback / comments if you write to the mailing list?
Thanks tsuna. I posted a note to the mailing list (https://groups.google.com/forum/?fromgroups=#!topic/opentsdb/ETTCexDDjNo%5B1-25-false%5D). Comments welcome!
BTW I would like to encourage you to name your repository something other than "opentsdb" unless it's actually a fork of opentsdb's repo, to avoid any confusion.
Thanks, good point. Done and done.
Thanks :)
Hi Holstius, I saw your work on opentsdbr. Since you are using HTTP API. There is performance issue with huge data. I am working similar stuff. But read data directly from hbase table 'tsdb' and plotting graph. I am using rbase/Rhipe, something similar to Rhadoop. I need to know how to get data from table 'tsdb' and 'tsdb-uid'. I don't have much idea on this table. how it is store. So can any one help me with hbase shell code. So I can carry up with rest stuff.
Looking forward for help
Hi @bikash,
you may want to read the documentation on OpenTSDB's schema to understand how to read the data in the tsdb
table.
Thank you @tsuna, I read the document, But since data store is in binary format. I don't how to make row key using tsdb-uid table.
Hi @holstius, After I finished the installation step: Install directly from GitHub using devtools for the opentsdbr package. I ran the first example, after the step: (result <- tsd_get(metric, start, tags=c(site="*"), downsample="10m-avg")) I got the error message: Error: is.interval(interval) is not TRUE Do you know the reason why I the error message? Thanks, Jian
https://github.com/holstius/opentsdbr
I think that issues for the library has to be asked in the opentsdr repo
@manolama You can close this
@matteoredaelli if your question has been answered, could you please close this issue ? thanks :)
Hello,
For the project Strategico (http://code.google.com/p/strategico/) I would like to save/read the time series to/from opentsdb.
I would like to implement a R (http://www.r-project.org/) library for using opentsdb using tcp/http protocols but I see that HTTP API (http://opentsdb.net/http-api.html ) (or its documentation) misses some features...
Is it possible to have a full TCP/HTTP interface? possibly http restful?
Thanks Matteo http://www.redaelli.org/matteo