Re: [PATCH proxmox-backup v8 00/17] partially fix #6563: add s3 counter for statistics and notifications

[Dominik Csapak <d.csapak@proxmox.com>]

didn't look too closely through the backend code yet, but wanted
to give some early feedback:

aside from the one comment on the summary page, the ui/ux looks
good to me.

I noticed that in my notifications, the datastore seems to be not set?
I get a notification with

Threshold Exceeded For Datastore ''

which seems wrong, but the code looks as if it should be set, so
maybe there is something going on with my setup though...

also the series needs a rebase, i applied it on an older commit
and rebased it myself (should be trivial to do from what it looks like)

On 4/2/26 12:53 PM, Christian Ebner wrote:
> This patch series implements request and traffic counters for the s3
> client, being shared as atomic counters via shared memory and mmapping
> across all s3-client instances.
> 
> The shared counters are instantiated on demand for individual
> datastores with s3 backend on s3 client instantiation and stored as
> part of the datastore implementation by loading the mmapped file on
> datastore creation, being cached for further access. Counter threshold
> values can be defined in the datastore options, allowing to send
> notifications via the notification system if counters exceeded the set
> threshold values. Further, a per-datastore schedule can be configured
> to reset the counters with given interval.
> 
> Further, counters are being loaded during rrd metrics collection and
> exposed as charts in the datastore summary, including the total
> request counts per method, the total upload and download traffic as
> well as the averaged upload and download rate.
> 
> Usage statistics, are not included in this series an will be tackled
> as followup series.
> 
> Link to the bugtracker issue:
> https://bugzilla.proxmox.com/show_bug.cgi?id=6563
> 
> Changes since version 7 (thanks a lot @Thomas for review and testing):
>   Adapt RRD graphs for S3 request counters to not fill area, have no
>    transparency and double the line width, making them more accessible
> - Drop patch for notification counter threshold rendering, inline to
>    callside instead
> - Display `None` if no notification threshold counters or reset
>    schedule is configured
> 
> Changes since version 6 (thanks a lot @Thomas for review and @Hannes
> for review and testing):
> - Return last counter values on reset and show them in the scheduled
>    task log (as suggested off-list by Thomas)
> - Add meaningful description/title for task list and task log
> - Directly use S3RequestCounterConfig instead of wrapping the options
>    previously stored in S3RequestCounterOptions in it
> - Drop useless Arc for callback fn as it can be shared without
> - Fix swapped comments in api types for upload/download traffic
> - Rebased onto current master
> 
> Changes since version 5 (thanks a lot @Hannes for review and testing):
> - Implemented the scheduled reset logic on top of the previous series
> - Also upload/download traffic counters are now reset, not just request counters
> - Fixed the possible swallowing of errors in `Content`s `Body` implementation
> - request_flush() is no longer called by spawning a blocking tokio task
> - Fixed the notification callback to not capture the datastore name, but provide
>    a `label` as callback parameter instead
> - Fix the api definition for the reset_counters endpoint
> - Fixed incorrect graph labels and typos discovered during review
> 
> Changes since version 4:
> - Rebase on top of patch series https://lore.proxmox.com/pbs-devel/20260311095906.202410-1-c.ebner@proxmox.com/T/
>    as requested by Fabian in order to facilitate review flow.
> 
> Changes since version 3:
> - Extend previous implementation by threshold values for counters and
>    mechanism to send notifications once thresholds are exceeded
> 
> Changes since version 2:
> - Introduce msync flushing mechanism to periodically persist counters
>    state to mmap backing file.
> 
> Changes since version 1 (thanks @Robert for feedback):
> - Keep tmpfs check in shmem mapping for pre-existing code, add
>    dedicated methods to create mmapped files in persistent locations.
> - Reorder and align atomic counters to 32-byte (half default cache
>    line size) to reduce false sharing.
> - Rework request counter init logic, avoid unsafe code and undefined
>    behaviour.
> - Rework page size calculation based on new counter alignment.
> - Avoid the need to open and mmap counter file for each rrd data
>    collection, keep the per-datastores filehandle cached instead.
> 
> 
> proxmox-backup:
> 
> Christian Ebner (17):
>    api: s3: add endpoint to reset s3 request counters
>    bin: s3: expose request counter reset method as cli command
>    ui: datastore summary: move store to be part of summary panel
>    ui: expose s3 request counter statistics in the datastore summary
>    metrics: collect s3 datastore statistics as rrd metrics
>    api: admin: expose s3 statistics in datastore rrd data
>    partially fix #6563: ui: expose s3 rrd charts in datastore summary
>    datastore: refactor datastore lookup parameters into dedicated type
>    api: config: update notification thresholds for config and counters
>    ui: add notification thresholds edit window
>    notification: define templates and template data for thresholds
>    datastore: add thresholds notification callback on datastore lookup
>    api/ui: notifications: add 'thresholds' as notification type value
>    api: config: allow counter reset schedule editing
>    ui: expose counter reset schedule edit window
>    bin: proxy: periodically schedule counter reset task
>    ui: add task description for scheduled counter reset
> 
>   debian/proxmox-backup-server.install          |   2 +
>   pbs-datastore/src/datastore.rs                |  93 ++++-
>   pbs-datastore/src/lib.rs                      |   3 +-
>   pbs-datastore/src/snapshot_reader.rs          |   7 +-
>   src/api2/admin/datastore.rs                   |  45 ++-
>   src/api2/admin/namespace.rs                   |   9 +-
>   src/api2/admin/s3.rs                          |  71 +++-
>   src/api2/backup/mod.rs                        |   3 +-
>   src/api2/config/datastore.rs                  |  42 ++-
>   src/api2/config/notifications/mod.rs          |   1 +
>   src/api2/reader/mod.rs                        |   3 +-
>   src/api2/status/mod.rs                        |   6 +-
>   src/api2/tape/backup.rs                       |   6 +-
>   src/api2/tape/restore.rs                      |   6 +-
>   src/bin/proxmox-backup-proxy.rs               |  89 ++++-
>   src/bin/proxmox_backup_manager/s3.rs          |  33 ++
>   src/server/metric_collection/metric_server.rs |   8 +-
>   src/server/metric_collection/mod.rs           |  90 ++++-
>   src/server/metric_collection/pull_metrics.rs  |  18 +-
>   src/server/metric_collection/rrd.rs           |  34 +-
>   src/server/notifications/mod.rs               |  39 +-
>   src/server/notifications/template_data.rs     |  30 ++
>   src/server/prune_job.rs                       |   3 +-
>   src/server/pull.rs                            |   7 +-
>   src/server/push.rs                            |   3 +-
>   src/server/verify_job.rs                      |   3 +-
>   src/tools/mod.rs                              |  14 +-
>   templates/Makefile                            |   2 +
>   .../default/thresholds-exceeded-body.txt.hbs  |   9 +
>   .../thresholds-exceeded-subject.txt.hbs       |   1 +
>   www/Makefile                                  |   2 +
>   www/Utils.js                                  |   5 +
>   www/datastore/OptionView.js                   |  16 +
>   www/datastore/Summary.js                      | 344 ++++++++++++------
>   www/window/CounterResetScheduleEdit.js        |  27 ++
>   www/window/NotificationThresholds.js          | 113 ++++++
>   36 files changed, 996 insertions(+), 191 deletions(-)
>   create mode 100644 templates/default/thresholds-exceeded-body.txt.hbs
>   create mode 100644 templates/default/thresholds-exceeded-subject.txt.hbs
>   create mode 100644 www/window/CounterResetScheduleEdit.js
>   create mode 100644 www/window/NotificationThresholds.js
> 
> 
> Summary over all repositories:
>    36 files changed, 996 insertions(+), 191 deletions(-)
>