[fix] Handle index db corruption and warn in UI

CHANGELOG.md

@@ -1,5 +1,8 @@
 # Changelog
 
+## 3.19.4
+- Handle index db corruption and warn in UI (mihran113) 
+
 ## 3.19.3 Apr 17, 2024
 - Resolve issue with new runs after tracking queue shutdown (mihran113)
 - Reset base path when opening new tabs (mihran113)

aim/storage/union.pyx

@@ -1,6 +1,8 @@
 import heapq
 import logging
 import os
+import shutil
+
 import aimrocks
 
 import cachetools.func
@@ -145,6 +147,8 @@ class ValuesIterator(ItemsIterator):
 
 
 class DB(object):
+    _corruption_warned = False
+
     def __init__(self, db_path: str, db_name: str, opts, read_only: bool = False):
         assert read_only
         self.db_path = db_path
@@ -181,6 +185,18 @@ class DB(object):
         index_path = os.path.join(self.db_path, self.db_name, "index")
         try:
             index_db = self._get_db(index_prefix, index_path, self._dbs)
+            # do a random read to check if index db is corrupted or not
+            index_db.get(index_prefix)
+        except aimrocks.errors.RocksIOError:
+            # delete index db and mark as corrupted
+            corruption_marker = Path(index_path) / '.corrupted'
+            if not corruption_marker.exists():
+                logger.warning('Corrupted index db. Deleting the index db to avoid errors. '
+                               'Please run `aim storage reindex command to restore optimal performance.`')
+                shutil.rmtree(index_path)
+                Path(index_path).mkdir()
+                corruption_marker.touch()
+            index_db = None
         except Exception:
             index_db = None
             logger.info('No index was detected')

aim/web/api/projects/pydantic_models.py

@@ -7,6 +7,7 @@ class ProjectApiOut(BaseModel):
     path: str
     description: str
     telemetry_enabled: int
+    warn_index: Optional[bool] = False
 
 
 class ProjectParamsOut(BaseModel):

aim/web/api/projects/views.py

@@ -1,7 +1,7 @@
 import os
 from datetime import timedelta
 from typing import Optional, Tuple
-
+from logging import getLogger
 from collections import Counter
 from fastapi import Depends, HTTPException, Query, Header
 from aim.web.api.utils import APIRouter  # wrapper for fastapi.APIRouter
@@ -21,6 +21,8 @@
 
 projects_router = APIRouter()
 
+logger = getLogger()
+
 
 @projects_router.get('/', response_model=ProjectApiOut)
 async def project_api():
@@ -29,11 +31,20 @@ async def project_api():
     if not project.exists():
         raise HTTPException(status_code=404)
 
+    # check if the index db was corrupted and deleted
+    corruption_marker = os.path.join(project.repo_path, 'meta', 'index', '.corrupted')
+    warning_message = ''
+    if os.path.exists(corruption_marker):
+        warning_message = 'Index db was corrupted and deleted. ' \
+                          'Please run `aim storage reindex` command to restore optimal performance.'
+        logger.warning(warning_message)
+
     return {
         'name': project.name,
         'path': project.path,
         'description': project.description,
         'telemetry_enabled': 0,
+        'warn_index': bool(warning_message),
     }
 
 

aim/web/ui/src/App.tsx

@@ -1,5 +1,6 @@
 import React from 'react';
 import { BrowserRouter, Switch, Route, Redirect } from 'react-router-dom';
+import { useModel } from 'hooks';
 
 import { loader } from '@monaco-editor/react';
 
@@ -16,8 +17,12 @@ import PageWrapper from 'pages/PageWrapper';
 
 import routes from 'routes/routes';
 
+import projectsModel from 'services/models/projects/projectsModel';
+
 import { inIframe } from 'utils/helper';
 
+import { IProjectsModelState } from './types/services/models/projects/projectsModel';
+
 import './App.scss';
 
 const basePath = getBasePath(false);
@@ -32,6 +37,7 @@ loader.config({
 });
 
 function App(): React.FunctionComponentElement<React.ReactNode> {
+  const projectsData = useModel<Partial<IProjectsModelState>>(projectsModel);
   React.useEffect(() => {
     let timeoutId: number;
     const preloader = document.getElementById('preload-spinner');
@@ -56,6 +62,12 @@ function App(): React.FunctionComponentElement<React.ReactNode> {
             <b>keep server running</b> for a better experience
           </AlertBanner>
         )}
+        {projectsData?.project?.warn_index && (
+          <AlertBanner type='warning'>
+            Index db was corrupted and deleted. Please run
+            <b>`aim storage reindex`</b> command to restore optimal performance.
+          </AlertBanner>
+        )}
         <div className='pageContainer'>
           <ErrorBoundary>
             <SideBar />

aim/web/ui/src/types/services/models/projects/projectsModel.d.ts

@@ -4,6 +4,7 @@ export interface IProject {
   name?: string;
   path?: string;
   telemetry_enabled?: string | boolean;
+  warn_index?: boolean;
 }
 
 export interface IProjectParamsMetrics {