Skip to content
CC Switch WikiUnofficial Community Wiki

Usage Stats Missing and Disk Usage at 100% - CC Switch

Quick fix

Most related issues remain unresolved. Updating to the latest version and disabling the local proxy may restore usage display and reduce disk I/O.

Symptom

Section titled “Symptom”
Error output
频繁sqlite写库，对hdd机械盘不友好，一启动ccswitch任务管理磁盘使用率就干满100%
English translation

Frequent SQLite writes are unfriendly to HDDs. As soon as cc-switch starts, the task manager shows disk usage maxed out at 100%.

Cause

Section titled “Cause”

This issue cluster includes several related symptoms. The core mechanism is that cc-switch's local proxy feature uses SQLite to record requests and session usage frequently. In v3.18.0, SQLite is configured with journal_mode=delete and synchronous=2(FULL); each small transaction triggers multiple fsync calls, causing about 250–350× write amplification on HDDs. In addition, the backend's 60-second session sync timer uses the default MissedTickBehavior::Burst. After Windows wakes from sleep, it sends every missed tick with no interval, scanning thousands of session log files in full and keeping CPU at 100% for a long time. The missing usage stats issue (#3138) was restored in v3.16.0, and the Codex usage stats issue (#7081) was fixed in PR #6905. The opencode_session request log's total duration is always 0.0s (#6326) because latency_ms is hardcoded to 0 in the source. Enabling the routing master switch also causes the WebView2 GPU process to continuously use 10%–20% GPU (#5760).

Fix

Section titled “Fix”

  1. Update cc-switch to the latest version. v3.16.0 restored the remaining quota display in the status bar and main window, and PR #6905 fixed the Codex usage stats issue.

  2. If you do not need the local proxy feature, turn off the local proxy master switch in cc-switch settings to reduce disk I/O and GPU usage immediately.

  3. If you use an HDD, move the .cc-switch database directory to an SSD, or add an SSD to the computer to ease the random-write IOPS bottleneck.

  4. For the issue where the opencode_session request log's total duration is always 0.0s, latency_ms is currently hardcoded to 0 in the source; wait for an official fix.

Affected Versions

Section titled “Affected Versions”
ToolClaude Code / Codex / OpenCode
Version3.15 - 3.20.1（受影响版本）
PlatformsWindowsmacOS

Source Issues

Section titled “Source Issues”

This page is distilled from 7 real issues

FAQ

Section titled “FAQ”
What should I do if the remaining quota is missing after an update?
Issue #3138 confirms that v3.16.0 restored the remaining quota display. Update to that version or later.
Why is GPU usage high after enabling routing?
Real-time refreshes in the routing panel and usage stats page continuously trigger WebView2 re-rendering, causing 10%–20% GPU usage while idle. Turn off the routing master switch to restore normal usage.
Why does disk usage hit 100% on an HDD?
The local proxy frequently logs request data through SQLite. journal_mode=delete + synchronous=FULL causes hundreds-fold write amplification on HDDs. Disabling the local proxy or moving the database to an SSD may help.
What should I do if CPU stays at 100% after Windows wakes from sleep?
The backend's 60-second session sync timer uses the Burst policy. After wake, it sends every missed tick, triggering a full scan. This issue (#5164) is still open; restarting cc-switch is a temporary workaround.
What should I do if Codex usage stats look wrong?
Issue #7081 confirms that this issue was fixed in PR #6905. Update to a version that includes the fix.
Why is the total duration for opencode_session request logs always 0.0s?
In session_usage_opencode.rs's insert_opencode_message, latency_ms is hardcoded to 0 and first_token_ms to None. This issue (#6326) is still open; wait for an official fix.
Section titled “Related problems”
Tray menu does not show current Plan allowance when multiple ChatGPT accounts are linked - CodexUpgrade cc-switch: the tray menu failing to show allowance with multiple linked ChatGPT accounts is a confirmed bug, fixed in commit e53ab21 and released…3 issues mergedQuery Failed: cc-switch GLM (Zhipu) usage queryFor GLM, "Query Failed" may only mean usage query is unavailable; API forwarding can still work. Upgrade and retest, or use a custom template.1 issues mergedCost always 0: GPT-6 Sol usage statistics in CodexEnable “Model Price Sync” in Settings to synchronize pricing data for gpt-6-sol.1 issues merged

This is an unofficial community wiki with no affiliation to the cc-switch authors or the project itself. Its content is compiled from the project's public GitHub issues. This site distributes no software.