-
Notifications
You must be signed in to change notification settings - Fork 40
FAQ
There currently is not a tool to migrate data from Harvest 1.6 to 2.0. The most common workaround is to run both, 1.6 and 2.0, in parallel. Run both, until the 1.6 data expires due to normal retention policy, and then fully cut over to 2.0.
Technically, it’s possible to take a Graphite DB, extract the data, and send it to a Prometheus db, but it’s not an area we’ve invested in. If you want to explore that option, check out the promtool which supports importing, but probably not worth the effort.
Is there a way to allow per SVM level user views? I need to offer 1 tenant per SVM. Can I limit visibility to specific SVMs? Is there an SVM dashboard available?
You can do this with Grafana. Harvest can provide the labels for SVMs. The pieces are there but need to be put together.
Grafana templates support the $__user variable to make pre-selections and decisions. You can use that + metadata mapping the user <-> SVM. With both of those you can build SVM specific dashboards.
There is a German service provider who is doing this. They have service managers responsible for a set of customers – and only want to see the data/dashboards of their corresponding customers.
What permissions does Harvest need to talk to ONTAP?
Permissions, authentication, role based security, and creating a Harvest user are covered here.
How do I make Harvest collect additional ONTAP counters?
Instead of modifying the out-of-the-box templates in the conf/ directory, it is better to create your own custom templates following these instructions.
How are capacity and other metrics calculated by Harvest?
Each collector has its own way of collecting and post-processing metrics. Check the documentation of each individual collector (usually under section #Metrics). Capacity and hardware-related metrics are collected by the Zapi collector which emits metrics as they are without any additional calculation. Performance metrics are collected by the ZapiPerf collector and the final values are calculated from the delta of two consequent polls.
How do I tag ONTAP volumes with metadata and surface that data in Harvest?
See volume tagging issue and volume tagging via sub-templates
How do I relate ONTAP REST endpoints to ZAPI APIs and attributes?
Please refer to the ONTAPI to REST API mapping document.
How much disk space is required by Prometheus?
This depends on the collectors you've added, # of nodes monitored, cardinality of labels, # instances, retention, ingest rate, etc. A good approximation is to curl your Harvest exporter and count the number of samples that it publishes and then feed that information into a Prometheus sizing formula.
Prometheus stores an average of 1-2 bytes per sample. To plan the capacity of a Prometheus server, you can use the rough formula: needed_disk_space = retention_time_seconds * ingested_samples_per_second * bytes_per_sample A rough approximation is outlined https://devops.stackexchange.com/questions/9298/how-to-calculate-disk-space-required-by-prometheus-v2-2
In Grafana, why do I see more results from topk than I asked for?
Topk is one of Prometheus's out-of-the-box aggregation operators, and is used to calculate the largest k elements by sample value.
Depending on the time range you select, Prometheus will often return more results than you asked for. That's because Prometheus is picking the topk for each time in the graph. In other words, different time series are the topk at different times in the graph. When you use a large duration, there are often many time series.
This is a limitation of Prometheus and can be mitigated by:
- reducing the time range to a smaller duration that includes fewer topk results - something like a five to ten minute range works well for most of Harvest's charts
- the panel's table shows only the topk rows and that data can be used to supplement the additional series shown in the charts