Who | Notes |
---|
| - File transfer kicked on on Friday, but from midnight (PST) to 2am it slows down, but it picks back up after 2am
- Files are being transferred individually
- throughput drops out, what is happening there? Discuss in Slack
- Up to 800 TB transferred so far.
- Location? Right now it's under scratch. How do we coordinate users?'
- Originally, same layout as NCSA. Perhaps have a "sandbox" environment?
- Will move the other directories.
|
| - Have messages from Rucio to Camel
- Multiple Kafka brokers, or all in one?
- Machine size?
- IN2P3 account available
- UK has experience with Kafka scaling that can prove helpful
|
| - Made progress with configuration testing with the storage element in front of Slurm
- doing isolated tests with the arc6 clients (will start with hello jobs to start)
- We will try to deploy an instance of FTS of why we're not transferring at speed with SLAC
- Might be file by file auth causing the slowdown
|
| Discussions with trying to run PanDA at SLAC |
| - We are stuck, MG has been helping, at the Hello World testing stage, submitting a command on PanDA on a test queue
- we think there's been progress with getting the Google logging to work
- not sure what next steps are, probably a meeting with SLAC
- Hsin-Fang has been working with Google Logging with Wen Guan
- Looking for best directory space to use for this.
- Will start a conversation with SLAC
- In France, we're using environment variables for the auth info and other info
- Suggest that we coordinate on how we'll do this between the sites.
- Re: Logging, when we'll have all the facilities working, we'll need (authorization) to push log entries into Google, and accounts so we can make sure it gets there (either dashboards/scripts, etc).
- K-T suggests that humans with accounts would probably be best.
- F: Scripts could use a service account
- MG: we should be sure we retain independence for the different sites, to retain the efficiencies at each site.
- F: Any examples?
- PanDA has some examples, but everyone is working on how we do this from SLAC
- BY: any PanDA available elsewhere?
- Richard: We're working on it, but it's still being set up.
|
| (previously discussed) |
| - Was working on CMS last week, but did one transfer request last week
|
| no updates |
General Discussion | - BW: Working on Rucio - need to start upgrading Rucio version to 1.27 and 1.28 (with an eye on 1.29). Lots of things to upgrade.
- Kustomize should be interesting for Rucio (which is built into the kubectl command)
- Argo - https://argo-cd.readthedocs.io/en/stable/
|
|
|