Registry & Domain
- Following up
Collectors
- Meeting about collectors took place with Jisc and GEANT
- ESnet planning to establish public Kafka cluster that could be used to inter-change data between R&E collectors
- We're going to test it with GEANT(eu.scitags.org) collector passing fireflies to ESnet
- Discussed operating model
- Looking for good use cases for network operations that would motivate Scitags deployment
- One such could be improving visibility of LHC traffic on GPN (nonLHC) as well as better granularity within LHCONE
- Correlation with network operations data might be another (sFlow, etc.)
- Ideas, suggestions and/or experiences are welcome
- Contacted by Andrew Elwell, currently working at Pawsey HPC in Australia
- They’re interested in Scitags to get better visibility into network flows, they’re running mainly astro-physics (SKA/AusSRC, ASKAP, MWA)
- We're looking for contacts in SKAO and AARNET (potentially to operate Asia/Pacific collector)
Collector Dashboards
- Topology: One of the re-occurring topics is ability to assign IP addresses to sites
- WLCG CRIC is good source but has only LHCONE/LHCOPN prefixes
- Looking for ways how we could collect all prefixes from all sites using Scitags
- Two new dashboards
- R&E dashboard - some minor changes and fixes, throughput calculation and topology needs some additional work
- Site dashboard - looks at performance of the storage system (pool nodes)
Experiments
- LHCb plans to review our requirements and implement Scitags in the next couple of weeks
dCache
- 10.2.13 testing at AGLT2. We found a possible memory leak that we need to track down and fix before testing can continue
FTS/XRootd (investigated missing fireflies at CERN collector)
- Backtested Xrootd implementation to compare 5.7.0 vs 5.8.0 and how it behaves wrt. HTTP headers
- Details at: https://docs.google.com/drawings/d/12vKqCK33LkgQxKwcayxxGRXcOcD6GOKrumx72jzb0m8/edit?usp=sharing
- We need to decide whether we want to be able to collect all traffic related to a specific site (at that site)
- This is important if we want to run site collectors, but also for overall coverage (to avoid requiring all sites to support scitags to get all data)
- To achieve this, we will need to request FTS to modify the HTTP headers in GET requests so that the source site issues the firefly instead of the destination.
flowd-go
- Meeting with Pablo at CERN, some updates on eBPF and netlink readout
- flowd-go containers are now public