nidx-binding 6.9.6.post673__tar.gz → 6.9.6.post677__tar.gz

{nidx_binding-6.9.6.post673 → nidx_binding-6.9.6.post677}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: nidx_binding
-Version: 6.9.6.post673
+Version: 6.9.6.post677
 Classifier: Programming Language :: Rust
 Classifier: Programming Language :: Python :: Implementation :: CPython
 Classifier: Programming Language :: Python :: Implementation :: PyPy

{nidx_binding-6.9.6.post673 → nidx_binding-6.9.6.post677}/nidx_protos/pyproject.toml

@@ -10,7 +10,7 @@ build-backend = "pdm.backend"
 
 [project]
 name = "nidx_protos"
-version = "6.9.6.post673"
+version = "6.9.6.post677"
 license = "AGPL-3.0-or-later"
 description = "Protobuf definitions for nucliadb/nidx"
 authors = [{ name = "Nuclia", email = "nucliadb@nuclia.com" }]

{nidx_binding-6.9.6.post673 → nidx_binding-6.9.6.post677}/pyproject.toml

@@ -13,7 +13,7 @@ classifiers = [
     "Programming Language :: Python :: Implementation :: CPython",
     "Programming Language :: Python :: Implementation :: PyPy",
 ]
-version = "6.9.6.post673"
+version = "6.9.6.post677"
 
 [project.urls]
 Homepage = "https://nuclia.com"

{nidx_binding-6.9.6.post673 → nidx_binding-6.9.6.post677}/src/control.rs

@@ -34,17 +34,24 @@ use crate::{NidxMetadata, metrics, settings::EnvSettings, telemetry};
 pub enum ControlRequest {
     Alive,
     Ready,
-    SetLogLevel { level: String },
+    SetLogLevel {
+        level: String,
+    },
+    CheckDatabase {
+        #[arg(action = clap::ArgAction::Set)]
+        enabled: bool,
+    },
 }
 
 #[derive(Debug, Serialize, Deserialize)]
 struct Alive {
     database: bool,
+    check_database: bool,
 }
 
 impl Alive {
     fn all_ok(&self) -> bool {
-        self.database
+        !self.check_database || self.database
     }
 }
 
@@ -79,6 +86,7 @@ pub struct ControlServer {
     meta: NidxMetadata,
     searcher_synced: Option<bool>,
     nats_client: Option<async_nats::Client>,
+    check_database: bool,
 }
 
 impl ControlServer {
@@ -87,6 +95,7 @@ impl ControlServer {
             meta,
             searcher_synced: has_searcher.then_some(false),
             nats_client,
+            check_database: true,
         }
     }
 
@@ -120,12 +129,24 @@ impl ControlServer {
             ControlRequest::Alive => ControlResponse::Alive(self.alive().await),
             ControlRequest::Ready => self.ready().await,
             ControlRequest::SetLogLevel { level } => telemetry::set_log_level(&level).into(),
+            ControlRequest::CheckDatabase { enabled } => {
+                self.check_database = enabled;
+                ControlResponse::Success
+            }
         }
     }
 
     async fn alive(&self) -> Alive {
-        Alive {
-            database: self.meta.pool.acquire().await.is_ok(),
+        if self.check_database {
+            Alive {
+                database: self.meta.pool.acquire().await.is_ok(),
+                check_database: true,
+            }
+        } else {
+            Alive {
+                database: true,
+                check_database: false,
+            }
         }
     }
 

{nidx_binding-6.9.6.post673 → nidx_binding-6.9.6.post677}/src/searcher/shard_selector.rs

@@ -26,6 +26,7 @@ use std::{
     hash::{DefaultHasher, Hash as _, Hasher as _},
     pin::pin,
     sync::Arc,
+    time::Duration,
 };
 use tokio_util::sync::CancellationToken;
 use tracing::*;
@@ -75,11 +76,21 @@ impl KubernetesCluster {
             // Read the stream continuously in order to keep the store updated
             let mut stream = pin!(stream);
             let mut prev_pods = HashSet::new();
+            let mut retries = 0;
             loop {
                 tokio::select! {
                     result = stream.try_next() => {
                         if let Err(e) = result {
-                            return Err(e.into())
+                            warn!(?e, "Error updating kubernetes cluster");
+                            if retries > 5 {
+                                return Err(e.into())
+                            } else {
+                                retries += 1;
+                                tokio::time::sleep(Duration::from_secs(5)).await;
+                                continue;
+                            }
+                        } else if retries > 0 {
+                            retries -= 1;
                         }
                         let new_pods = task_reader
                             .state()