aboutsummaryrefslogtreecommitdiffstats
path: root/node-repository/src/main/java/com/yahoo/vespa/hosted/provision/node/Report.java
Commit message (Collapse)AuthorAgeFilesLines
* Update copyrightJon Bratseth2023-10-091-1/+1
|
* Update 2019 Oath copyrights.gjoranv2021-10-271-1/+1
|
* Remove "encrypt" decommission operationMartin Polden2021-10-151-5/+0
|
* Trigger host restart when encryptingMartin Polden2021-05-281-0/+1
|
* Implement HostEncrypterMartin Polden2021-05-271-0/+4
|
* Preserve only specific fields when reprovisioningJon Bratseth2020-03-201-0/+3
|
* Move SlimeUtils to vespajlibMartin Polden2020-02-061-1/+1
|
* Use the type of the node reportHåkon Hallingstad2019-02-281-11/+49
| | | | | | Preserve the type of the node reports. Use the type to make decisions in NodeFailer and FailedExpirer.
* Retire/fail hosts with failure reportsHåkon Hallingstad2019-02-121-5/+17
| | | | | | | | | | NodeFailer will: - fail a ready node if the host has failure reports. - set wantToRetire on an active parent host, and all children, if it has failure reports. Makes the NodeFailer operate as a NodeFailer agent to make that clearer in the history.
* Support node reports in node repositoryHåkon Hallingstad2019-01-301-0/+74
Adds a "reports" field of the node JSON: - A JSON object mapping report IDs (e.g. cpuCyclesInGHz) to the report. - Supports GET and PATCH. Setting a report to null clears it. Setting "reports" to null clears all reports. Setting an already existing report will overwrite the previous. A report is of the form: { // The time the report was created/PATCHed, in millis since epoch (long). // If missing from the PATCH, it is set to now. "createdMillis": 123456, // Optional description. If set the NodeFailer should fail the node (for // now). "description": "Frequency...", // opaque additional data determined by the exact report (ID). // Can be queried by e.g. the NodeFailer, is saved in ZK, and will be // displayed as part of the report under "reports". // // Most hardware divergence reports will get a '"value": 1234' field with // the actually measured value (which was outside the threshold). ... } The host admin will be changed to upload to "reports" instead of hardware divergence. The NodeFailer will then be able to query the reports to decide which nodes, if any, to fail out.