1. EMI INFSO-RI-261611 Catalogue synchronization & ACL propagation Fabrizio Furano (CERN IT-GT-DMS)

2. 2 EMI INFSO-RI-261611 11 Apr 2011 F.Furano - Catalogue Synchronization & ACL propagation Various catalogues keep information that is related – E.g. LFC keeps info about the content of remote Storage Elements, each one with its own catalogue A change in the permissions of a file in LFC is not automatically reflected by the peripheric catalogue If a SE looses a file, the LFC does not know If a new file is not correctly registered -> dark data Keeping them in sync is a very hard problem Namespace scanning for diffs is an expensive workaround The problem

3. 3 EMI INFSO-RI-261611 11 Apr 2011 F.Furano - Catalogue Synchronization & ACL propagation Make the various catalogues/SE able to talk to each other – In order to exchange messages that keep them synchronized – 2 directions: Central Catalogue->SE (downstream) – e.g. to propagate changes in the permissions SE->Central Catalogue (upstream) – e.g. to propagate info about lost and missing files The idea

4. 4 EMI INFSO-RI-261611 11 Apr 2011 F.Furano - Catalogue Synchronization & ACL propagation Looking for good ways to reliably communicate and cooperate Communication SE1 SEn SE2 SE or exp. catalogue subscribes to the relevant topics (e.g. “Changes”) SE sends to the appropriate topics (e.g. “Lost”) Subscribes to the relevant topics (e.g. “Lost”) SE Sends to the appropriate topics (e.g. “Changes”) Central Catalogue Other catalogue/SE e.g. ATLAS Broker(s)

5. 5 EMI INFSO-RI-261611 11 Apr 2011 F.Furano - Catalogue Synchronization & ACL propagation Types of interactions What can we do with this? – Fix inconsistencies as they are found “SE1 apparently lost file X” – Prevent inconsistencies by sending messages when something happens “File X has new access permissions” “SE1 has a new file Y” – Ev. allow a central system to query the others to synchronize itself “Who has file Z”? “Do you still have file W?”

6. 6 EMI INFSO-RI-261611 11 Apr 2011 F.Furano - Catalogue Synchronization & ACL propagation The architecture Central Catalogue Central Catalogue Msg brokers Upstream topic “SEMsg_upstream” NotAvailable(sfn) [ FileCreated(sfn) ] Adapter Downstream topic “SEMsg_downstream” Chmod(sfn) SE or other Catalogue A file can be N/A if: - it was requested to a DB that does not know it - OR if it was requested to a GridFTP that does not find it (trickier) Fix info! Adapter Fix info! Uses a virtual destination, e.g. : Consumer.DPNS1.SEMsg_downstream the broker queues messages for this endpoint if it disconnects momentarily. Uses a virtual destination, e.g. : Consumer.LFC1.SEMsg_upstream the broker queues messages for this endpoint if it disconnects momentarily.

7. 7 EMI INFSO-RI-261611 11 Apr 2011 F.Furano - Catalogue Synchronization & ACL propagation SEMsg The current implementation is a component called SEMsg – Built to be robust, efficient and easy to integrate – Plugin-based (ev. with “null” plugins), loaded at runtime (dlopen) A plugin that performs actions (in the catalogue) when a message comes A plugin that performs SE(Catalogue)-specific actions when a message has to be sent through the API 5 plugins available by now: LFC producer+consumer, DPM producer+consumer, Python consumer – Provides a configurable daemon that consumes and dispatches the notifications – Provides commands to manually send notifications As well as a simple C/C++ API to send them hides message composition and the security implementation crafted to avoid bringing in complex build dependencies – The same tools are used for the LFC and DPM prototype, loading different sets of plugins Hence, more sets of plugins can be added, to talk to other systems

8. 8 Security DPM consumer plugin DPM consumer plugin Default producer plugin Default producer plugin Producer API, e.g. NotifyChmod(sfn) NotifyNotAvailable(sfn) Producer API, e.g. NotifyChmod(sfn) NotifyNotAvailable(sfn) LFC EMI INFSO-RI-261611 11 Apr 2011 F.Furano - Catalogue Synchronization & ACL propagation Detail - SEMsg plugins Security LFC consumer plugin LFC consumer plugin LFC producer plugin LFC producer plugin Producer API, e.g. NotifyChmod(sfn) NotifyNotAvailable(sfn) Producer API, e.g. NotifyChmod(sfn) NotifyNotAvailable(sfn) DPM Msg brokers These implement some kind of realtime storage aggregation Must be robust, fast and low latency To be able to trust them and build an evolving thing

9. 9 EMI INFSO-RI-261611 11 Apr 2011 F.Furano - Catalogue Synchronization & ACL propagation SEMsg Encapsulates the message-related aspects into a clear subsystem Simple to use and integrate in other systems Completely asynchronous, multithreaded design Does everything (in background) to keep the consumer connection UP The API never hiccups in case of conn troubles/broker restart The consumer may live as a daemon or be started within another daemon LFC/DPM use the external daemon Also provides commands to send the notifications – E.g. to manually notify that a set of files is not available Natively supports the bulk operations defined in the protocol – Automatic creation of bulk notifications from the internal queues. – No need for weird APIs or complex implementations.

10. 10 EMI INFSO-RI-261611 11 Apr 2011 F.Furano - Catalogue Synchronization & ACL propagation Security (1/2) We can: – Guarantee the identity of the senders – Make sure that the content has not changed – Support SSL connections to the brokers (=comm encryption) Everything is X509-based, described in detail in the docs Performance is good (hundreds/s), space for improvement, probably not needed now – The bulk messages help a lot here The messages are signed using SHA1 – We treat messages as we do with PGP for e-mails Future: implement some form of whitelisting

11. 11 EMI INFSO-RI-261611 11 Apr 2011 F.Furano - Catalogue Synchronization & ACL propagation Security (2/2) The communicating entities are in general SEs, hence machines running with: – a host certificate – a service certificate Now we have a way to be sure about the sender’s identity – The notifications are never destructive by construction – We might like to have a way to filter senders – Whitelisting seems a way to go – We need anyway some real-world experience

12. 12 EMI INFSO-RI-261611 11 Apr 2011 F.Furano - Catalogue Synchronization & ACL propagation New entry: the Python plugin One more consumer plugin, in the SEMsg distribution, with the DPM and LFC ones Associates python funcs to SEMsg notifications – Fully configurable in the SEMsg config file and generic e.g. The notification FileNotAvailable invokes the function ‘func1’ from the module ‘module1’ passing its content as parameters – Fast: invokes natively the Python C API – Benefits from the SEMsg structure, e.g. the security The Python script only deals with the action to be performed Tested with Python 2.4, 2.5, 2.6

13. 13 EMI INFSO-RI-261611 11 Apr 2011 F.Furano - Catalogue Synchronization & ACL propagation Upcoming: ATLAS DDM The first usage of the Python plugin is going to be to invoke the ATLAS DDM tools – The “DDM problematic file catalogue” can be directly fed with notifications coming from SEMsg, i.e. coming from: SEs, when they detect a missing file Manual invokation of the CLI with lists of files – In practice, an instance of the SEMsgdaemon will sniff the notifications and forward them through the Python plugin

14. 14 EMI INFSO-RI-261611 11 Apr 2011 F.Furano - Catalogue Synchronization & ACL propagation The Dec demonstrator was OK Now the libs/tools are available in the glite and EMI builds as a pre-release – All interested catalogues can use it – Documentation available in the TWiki, including the plugins https://twiki.cern.ch/twiki/bin/view/EMI/EmiJra1Syncat – The instrumentation of LFC/DPM is finding its way to the SVN :-) Installed in the DPM nightly testbed – Just a matter of installing a couple of RPMs from ETICS and setting a few options v1.1.0beta is available

15. 15 EMI INFSO-RI-261611 11 Apr 2011 F.Furano - Catalogue Synchronization & ACL propagation What’s next Sync with the other SE developers (STORM, dCache) Sync with LCGDM and provide the LFC/DPM instrumentation Start deploying the ATLAS instance and some real world test instances of LFC/DPM – Robustness tests + ev. fixes/additions in SEMsg/DPM/LFC At each step, keep an eye on the applicability to the computing models + ev. fixes/additions Would be really useful to understand if the protocol is descriptive enough Then, we’ll drop the ‘beta’

16. 16 EMI INFSO-RI-261611 11 Apr 2011 F.Furano - Catalogue Synchronization & ACL propagation Making catalogues and SEs interact seems a good way to attack the consistency problem – It’s a form of realtime interaction between SEs and catalogues – By definition, it won’t mathematically kill the inconsistencies, but will help making a much better system – Will be interesting to explore the possibilities of the technology and of the implementation SEMsg is available as a pre-release until EMI-2 Protocol and SEMsg documentation in the Wiki https://twiki.cern.ch/twiki/bin/view/EMI/EmiJra1Syncat Feedbacks are welcome The messaging (test) infrastructure and the tools seem really OK Conclusions

17. 17 EMI INFSO-RI-261611 11 Apr 2011 F.Furano - Catalogue Synchronization & ACL propagation Thank you EMI is partially funded by the European Commission under Grant Agreement INFSO-RI-261611