chore(cu): write wasm memory to tmp file then rename to prevent corru…

…ption of previous file on error

servers/cu/src/bootstrap.js

@@ -1,7 +1,7 @@
 import { dirname, join } from 'node:path'
 import { fileURLToPath } from 'node:url'
 import { randomBytes } from 'node:crypto'
-import { writeFile, mkdir } from 'node:fs/promises'
+import { writeFile, mkdir, rename as renameFile } from 'node:fs/promises'
 import { createReadStream } from 'node:fs'
 import { BroadcastChannel } from 'node:worker_threads'
 
@@ -172,6 +172,7 @@ export const createApis = async (ctx) => {
   const writeProcessMemoryFile = AoProcessClient.writeProcessMemoryFileWith({
     DIR: ctx.PROCESS_MEMORY_CACHE_FILE_DIR,
     writeFile,
+    renameFile,
     mkdir
   })
 
@@ -188,6 +189,7 @@ export const createApis = async (ctx) => {
      */
     DIR: ctx.PROCESS_MEMORY_FILE_CHECKPOINTS_DIR,
     writeFile,
+    renameFile,
     mkdir
   })
 

servers/cu/src/effects/ao-process.js

@@ -472,17 +472,26 @@ export function readProcessMemoryFileWith ({ DIR, readFile }) {
   }
 }
 
-export function writeProcessMemoryFileWith ({ DIR, writeFile, mkdir }) {
+export function writeProcessMemoryFileWith ({ DIR, writeFile, renameFile, mkdir }) {
   return ({ Memory, evaluation }) => {
-    const file = `state-${evaluation.processId}.dat`
+    const dest = `state-${evaluation.processId}.dat`
+    const tmp = `${dest}.tmp`
 
-    return lock.acquire(file, () => {
-      const { stop: stopTimer } = timer('writeProcessMemoryFile', { file })
-      const path = join(DIR, file)
+    return lock.acquire(dest, () => {
+      const { stop: stopTimer } = timer('writeProcessMemoryFile', { file: tmp })
 
+      const tmpPath = join(DIR, tmp)
+      const destPath = join(DIR, dest)
+      /**
+       * By first writing to a temp file, we prevent corrupting a previous file checkpoint
+       * if for whatever reason the writing is not successful ie. node process is terminated
+       *
+       * Then we rename the tmp to the dest file, effectively overwriting it, atomically.
+       */
       return mkdir(DIR, { recursive: true })
-        .then(() => writeFile(path, Memory))
-        .then(() => file)
+        .then(() => writeFile(tmpPath, Memory))
+        .then(() => renameFile(tmpPath, destPath))
+        .then(() => dest)
         .finally(stopTimer)
     })
   }