settingswatcher: call a function upon updates to TenantReadOnly settings #111210
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
This change is |
This was referenced Sep 25, 2023
knz
force-pushed
the
20230925-settings-watcher-callback
branch
2 times, most recently
from
72ad4fa to
774bb57
Compare
yuzefovich
approved these changes
Sep 25, 2023
There was a problem hiding this comment.
Reviewed 3 of 3 files at r13, all commit messages.
Reviewable status:complete! 1 of 0 LGTMs obtained
knz
force-pushed
the
20230925-settings-watcher-callback
branch
4 times, most recently
from
16ca4ff to
296384a
Compare
(This function/method will also be deleted in a later commit.) Release note: None
knz
force-pushed
the
20230925-settings-watcher-callback
branch
from
296384a to
5ec8c79
Compare
…ed cache (For context, on each node there is a local persisted cache of cluster setting customizations. This exists to ensure that configured values can be used even before a node has fully started up and can start reading customizations from `system.settings`.) Prior to this patch, entries were never evicted from the local persisted cache: when a cluster setting was reset, any previously saved entry in the cache would remain there. This is a very old bug, which was long hidden and was recently revealed when commit 2f5d717 was merged. In a nutshell, before this recent commit the code responsible to load the entries from the cache didn't fully work and so the stale entries were never restored from the cache. That commit fixed the loader code, and so the stale entries became active, which made the old bug visible. To fix the old bug, this present commit modifies the settings watcher to preserve KV deletion events, and propagates them to the persisted cache. (There is no release note because there is no user-facing release where the bug was visible.) Release note: None Co-authored-by: Steven Danna <[email protected]>
Prior to this patch, the rangefeed watcher over `system.settings` was updating the in-RAM value store before it propagated the updates to the persisted local cache. In fact, the update to the persisted local cache was lagging quite a bit behind, because the rangefeed watcher would buffer updates and only flush them after a while. As a result, the following sequence was possible: 1. client updates a cluster setting. 2. server is immediately shut down. The persisted cache has not been updated yet. 3. server is restarted. For a short while (until the settings watcher has caught up), the old version of the setting remains active. This recall of ghost values of a setting was simply a bug. This patch fixes that, by ensuring that the persisted cache is written through before the in-RAM value store. By doing this, we give up on batching updates to the persisted local store. This is deemed acceptable because cluster settings are not updated frequently. Release note: None
In a later commit, we will want to feed changes to TenantReadOnly settings in the system tenant's `system.settings` table into the connector, for use by the virtual cluster servers. This commit sets up the first building block: an extra callback function in the settings watcher that gets called when the watcher detects a change to TenantReadOnly settings. Release note: None
knz
force-pushed
the
20230925-settings-watcher-callback
branch
from
5ec8c79 to
d632d44
Compare
craig bot
pushed a commit
that referenced
this pull request
Sep 29, 2023
110758: server,settings: properly cascade defaults for TenantReadOnly r=stevendanna,yuzefovich a=knz Previous in sequence: - [x] #110789 and #110947 - [x] #110676 - [x] #111008 - [x] #111145 - [x] #111147 - [x] #111149 - [x] #111150 - [x] #111153 - [x] #111475 - [x] #111210 - [x] #111212 Fixes #108677. Fixes #85729. Fixes #91825. Completes the work described in the settings RFC. Epic: CRDB-6671 TLDR: this patch ensures that virtual cluster servers observe changes made to TenantReadOnly settings via SET CLUSTER SETTING in the system interface, even when there is no override set via ALTER VIRTUAL CLUSTER SET CLUSTER SETTING. For example, after `SET CLUSTER SETTING kv.closed_timestamp.target_duration = '10s'` in the system interface, this value will show up via `SHOW CLUSTER SETTING` in a virtual cluster SQL session. This changes the way that settings are picked up in virtual cluster, as follows: 1. if there is an override specifically for this tenant's ID (in `tenant_settings`), use that. 2. otherwise, if there is an override for the pseudo-ID 0 (in `tenant_settings` still, set via `ALTER TENANT ALL SET CLUSTER SETTING`), then use that. 3. **NEW** otherwise, if the class is TenantReadOnly and there is a custom value in `system.settings`, set via a regular `SET CLUSTER SETTING` in the system tenant, then use that. 4. otherwise, use the global default set via the setting's `Register()` call. ---- Prior to this patch, TenantReadOnly settings as observed from virtual clusters were defined as the following priority order: 1. if there is an override specifically for this tenant's ID (in `tenant_settings`), use that. 2. otherwise, if there is an override for the pseudo-ID 0 (in `tenant_settings` still, set via `ALTER TENANT ALL SET CLUSTER SETTING`), then use that. 3. otherwise, use the global default set via the setting's `Register()` call. Remarkably, this did not pick up any changes made via a plain `SET CLUSTER SETTING` statement via the system interface, which only modifies this setting's value in `system.settings` (thus not `tenant_settings`). This situation was problematic in two ways. To start, settings like `kv.closed_timestamp.target_duration` cannot be set solely in `system.tenant_settings`; they are also used in the storage layer and so must also be picked up from changes in `system.settings`. For these settings, it is common for operators to just issue the plain `SET CLUSTER SETTING` statement (to update `system.settings`) and simply forget to _also_ run `ALTER TENANT ALL SET CLUSTER SETTING`. This mistake is nearly unavoidable and would result in incoherent behavior, where the storage layer would use the customized value and virtual clusters would use the registered global default. The second problem is in mixed-version configurations, where the storage layer runs version N+1 and the SQL service runs version N of the executable. If the registered global default changes from version N to N+1, the SQL service would not properly pick up the new default defined in version N+1 of the storage layer. This patch fixes both problems as follows: - it integrates changes to TenantReadOnly settings observed in `system.settings`, to the watcher that tracks changes to `system.tenant_settings`. When a TenantReadOnly setting is present in the former but not the latter, a synthetic override is added. - it also initializes synthetic overrides for all the TenantReadOnly settings upon server initialization, from the registered global default, so that virtual cluster servers always pick up the storage layer's default as override. 111383: *: simplify tests r=yuzefovich a=knz All commits except the last are from #110758. Epic: CRDB-6671 Now that "tenant-ro" settings take their default from the system tenant's value, we do not need `ALTER TENANT ALL` for them any more. This patch simplifies test code accordingly. Rcommended by `@yuzefovich` in the review for #110758. 111440: cluster-ui: fix db page stories r=THardy98 a=THardy98 Epic: none This change fixes the stories for the database pages. Release note: None 111512: kv: correctly handle shared lock replays in KVNemesis r=nvanbenschoten a=arulajmani Previously, we'd return an AssertionFailed error if a SKIP LOCKED request discovered another request from its own transaction waiting in a lock's wait queue. In SQL's use of KV, this can only happen if the SKIP LOCKED request is being replayed -- so returning an error here is fine. However, this tripped KVNemesis up. This patch marks such errors, for the benefit of KVNemesis, and doesn't call them assertion failed errors. Fixes #111426 Fixes #111506 Fixes #111513 Release note: None Co-authored-by: Raphael 'kena' Poss <[email protected]> Co-authored-by: Thomas Hardy <[email protected]> Co-authored-by: Arul Ajmani <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
See the last commit.
Previous in sequence:
.Overrideset the value origin #111150In a later commit, we will want to feed changes to TenantReadOnly
settings in the system tenant's
system.settingstable into theconnector, for use by the virtual cluster servers.
This commit sets up the first building block: an extra callback
function in the settings watcher that gets called when the watcher
detects a change to TenantReadOnly settings.
Needed for #110758
Epic: CRDB-6671