> On Dec. 4, 2012, 7:29 p.m., Prasanna Santhanam wrote: > > I like the idea that this script can be provided as a convenience to > > collect various logs. One problem however is that the script assumes root > > access on the management server and executes a bunch of commands (ip > > routes, hostnames, top, df, dmesg) etc that someone might not be okay with. > > May be it should be made interactive and warn the user of the actions it > > will/has performed. When as a commercial support provider you have > > permission to access and troubleshoot a system it would be okay to run > > this. But IMO it could be provided by that company which is looking to > > provide support to the operator of the said cloud to simplify their process > > of support. So I'm slightly wary of accepting something like this unless > > someone else convinces me that this would absolutely be beneficial. Also > > admins operating large clouds might be using a syslog server to already do > > some (or more) of these actions for them. If the logs are moved away from > > where one expects them to be then this script fails. > > Radoslaw Smigielski wrote: > > So I'm slightly wary of accepting something like this unless someone > else convinces > > me that this would absolutely be beneficial. > Prasanna, let me try then. > > The main reasons why I created this tools: > 1. CloudStack is a project which targets cloud enthusiasts but also > "enterprises" and if you have a look on all the big products, vendors on the > marker they offer utilities which do exactly what cs-bugtool does. Collect > logs, basic system and configuration information. Examples: NetApp > autosupport, XenServer xs-bugtool, VMware vSphere Client does this, . . ., > and many others which I don't know. So IMHO CloudStack should also include > this type of utility. > 2. CloudStack becomes more and more complex, having this log bundle we > can share it with other trusted persons or some support representative and > this really can speed up analysis, avoid tens of questions, sending emails > and files back and forth. > 3. I fully understand your concern about sharing sensitive information > and to address this: > a.) I am going to implement some switches which let user decided what to > include in output archive > b.) It's clearly written in README but we can repeat this information > what exactly is collected > c.) We can add warning, the output archive can contain information which > you can consider as a sensitive please be aware of this and be careful with > who you sharing this(!!!) > d.) The main use case of this utility is to share output archive with > trusted people or some sort of support representative not to make public > users' configurations. We do not send any information automatically, we just > give a user a blob and user decides what to do with this. > > > script assumes root access on the management server and executes a > bunch of commands > This is a good point, I need to add logic which detect non-root execution > situation. > We this also should be in README. > > I hope above change your mind :) > > Radek. > > > Radoslaw Smigielski wrote: > > The collated logs may contain private information, and if you are at all > > worried about that, you should not use this tool, or you should > explicitly > > exclude those logs from the archive. > Prasanna, this is fragment of README.xen-bugtool taken from xen.org > project. > We can add similar disclaimer in our README and in cs-bugtool output. > > Radek. > > > Hugo Trippaers wrote: > I agree with the worry of Prasanna that this script is mainly useful for > parties that provide support professionally. That said it can not hurt to > have the tool in the main code, but it might trigger people to make these > blobs and send them to the -users mailing list looking for support. Not sure > yet if that is a good thing or not. A lot of deployments will have slight > changes based on the preferences of the administrators, so the script should > handle this gracefully. > > Note on testing, this should be tested on at least a recent ASF release > (4.0.0-incubating) or a recent branch (4.0 branch or master branch).
The idea is that a lot of stuff that has nothing to do with CloudStack directly should not be part of it. I would suggest that you work on this on your own git repo, say on github and share with users. Get their feedback on ML, implement new features and if everyone starts using this for sharing logs, we can go ahead and merge it. I only worry that if this gets committed now, and not used it would add bloat. I would really want to use this tool if this could also go to all the hosts and maybe ssvm/cpvm and get me logs, it would be awesome. But, if the folks don't give this a "ship it" I would want you to show 'em why we would want this tool with the next iteration of this tool and features. It makes sense to not commit tools that should n't be part of CS, for ex. let me give my personal examples; Initially I was writing cloudmonkey as a separate repo, but then I thought it is dependent on apis, marvin, so I moved it in. I've another cmd tool that helps me review, https://github.com/bhaisaab/RBTool I did not commit it because, even though I think it's an awesome tool to reviewing stuff and it was helpful to me at least during the 4.0 release. See repos by tsp, https://github.com/vogxn, he has a lot of 'em on test infra etc. but it does not make sense to move all of that code to CS-git. - Rohit ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/8248/#review14023 ----------------------------------------------------------- On Dec. 4, 2012, 6:49 p.m., Radoslaw Smigielski wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/8248/ > ----------------------------------------------------------- > > (Updated Dec. 4, 2012, 6:49 p.m.) > > > Review request for cloudstack, Prasanna Santhanam and Rohit Yadav. > > > Description > ------- > > The main purpose behind cs-bugtool is to make easier collecting > logs required for CloudStack/CloudPlatform troubleshooting > and it mainly comes from Support. cs-bugtool collects logs, system > information and cloud database dump, compress everything and creates > .zip archive file ready to share with others. > For those familiar with Xen/XenServer cs-bugtool name may sound similar > to xen-bugtool, this similarity is intentional. > > Current version of cs-bugtool collects: > - basic system and environment information > - CP/CS management service logs > - basic system logs > - cloud database > > Current version does not accept any arguments but this may > changed in next iterations of cs-bugtool. > > We DO NOT collect and do not want to collect any sensitive information. > If there is still any sensitive pice of information please let us know > and we will try to remove it or replace by meaningless data. > > > Diffs > ----- > > tools/support/README PRE-CREATION > tools/support/cs-bugtool PRE-CREATION > > Diff: https://reviews.apache.org/r/8248/diff/ > > > Testing > ------- > > I tested this on CS 3.0.4 and 3.0.5. > > > Thanks, > > Radoslaw Smigielski > >
