Hey
I just wanted to check something in Grafana and by that found out that my influxdb addon (addon in HA) is not working anymore.
No real clue when this started … .but at least a few weeks back it was working fine
Error says: Accept error: accept tcp [::]:8086: accept4: too many open files; retrying in 1s
2025/09/06 09:39:03 http: Accept error: accept tcp [::]:8086: accept4: too many open files; retrying in 1s
any tip what I could try is much appreciated
Thanks
Oh I found this is an open issue for the infludb addon
I hope @frenck will see this 
i have the same issue, and I think there are dozen more, they just do not check the influx data reguraly… it is a pittty that no solution is yet provided for it…
#!/bin/sh
IMAGE="ghcr.io/hassio-addons/influxdb/amd64:5.0.2"
ADDON="a0d7b954_influxdb"
echo "== Pull clean original image =="
docker pull "$IMAGE"
echo "== Create Dockerfile patch =="
rm -rf /mnt/data/influxdb-image-patch
mkdir -p /mnt/data/influxdb-image-patch
cat > /mnt/data/influxdb-image-patch/Dockerfile <<'DOCKER'
FROM ghcr.io/hassio-addons/influxdb/amd64:5.0.2
RUN if [ -f /etc/services.d/influxdb/run ]; then \
sed -i '/ulimit -S -n 65536/d' /etc/services.d/influxdb/run && \
sed -i '/ulimit -Sn 65536/d' /etc/services.d/influxdb/run && \
sed -i '/exec influxd/i ulimit -S -n 65536' /etc/services.d/influxdb/run && \
grep -n 'ulimit\|exec influxd' /etc/services.d/influxdb/run; \
else \
echo "ERROR: /etc/services.d/influxdb/run not found" && exit 1; \
fi
DOCKER
echo "== Build patched image using same local tag =="
docker build -t "$IMAGE" /mnt/data/influxdb-image-patch
echo "== Restart InfluxDB add-on =="
ha addons restart "$ADDON"
echo "== Wait =="
sleep 25
echo "== Verify =="
docker exec "addon_$ADDON" sh -c '
pid=$(pidof influxd || true)
if [ -z "$pid" ]; then
echo "ERROR: influxd not running"
exit 1
fi
echo "PID=$pid"
cat /proc/$pid/limits | grep "open files"
echo FD_COUNT
ls /proc/$pid/fd | wc -l
'
echo "== Done =="
I temporary fixed it by running this quick and dirty hotfix script (provided by chatgpt...) under Home Assistant Operating System 17.3 to be able to run newest kernel with fix for copy fail exploit.
Maybe it help you until real fix for the addon is there 
Make backup. Use on your own rist 
I had the same problem and used Claude to get me out of the mess. It took a lot of iterations back and forth, but in the end I succeeded. Here’s Claude’s final procedure:
InfluxDB add-on fails to start and the log shows repeated errors like:
Cannot read corrupt tsm file, renaming ... error="init: read tombstones: too many open files"
http: Accept error: accept tcp 0.0.0.0:8086: accept4: too many open files; retrying in 5ms
Starting with Home Assistant OS 16, the default file descriptor (open files) limit for
add-on containers was lowered to 1024. This is too low for InfluxDB, which needs to open
many files simultaneously on startup — one for each TSM data file, WAL file, and shard.
Over time, InfluxDB's internal monitoring database (_internal) accumulates hundreds of
shard folders, each containing multiple files. Combined with the lowered HAOS 16 limit,
InfluxDB can no longer open all required files on startup and fails.
Note: PR #395 on the hassio-addons/addon-influxdb GitHub repository proposes a proper
fix by making the file handler limit user-configurable. As of add-on version 5.0.2 this
fix has not yet been released.
The workaround in this guide deletes the _internal database, which only stores
InfluxDB's own internal metrics and has no value for Home Assistant users. This reduces
the number of files InfluxDB needs to open on startup back below the 1024 limit.
This unlocks additional add-ons in the store, including the Advanced SSH & Web Terminal
add-on needed for this procedure.
If not already installed:
This is required to get Docker access:
Remember to turn protection mode back on after you're done.
InfluxDB needs to be running for the next step. Even though it's broken, the container
still starts:
Click Advanced SSH in the HA sidebar. You'll see a terminal with a prompt like:
➜ ~
docker ps -a | grep influx
You should see a line with addon_XXXXXXXXX_influxdb. If you see it, proceed.
Run this command immediately after InfluxDB starts (before it fully loads):
docker exec addon_XXXXXXXXX_influxdb rm -rf /data/influxdb/data/_internal
Replace XXXXXXXXX with the actual add-on ID from the previous step.
docker exec addon_XXXXXXXXX_influxdb ls /data/influxdb/data/
You should see only your data folder (e.g. Homeassistant) and no _internal folder.
Check the log — it should now start cleanly with no too many open files errors.
Make sure the internal monitoring store is disabled:
falseThis prevents _internal from ever being recreated.
Direct browser access to Chronograf (port 8888) may not work. Instead, open the
InfluxDB web UI from within Home Assistant:
Homeassistant)_internal folder (InfluxDB's own internal metrics) is removedFyi to you, Claude & anyone else reading this, advanced mode was removed as of 2026.6 release 2 days ago & is now the default.
Skip this step if you're on the latest release & always tell Claude to verify against the latest HA docs/ release notes.
For anybody, to whom the provided solution does not work (deleting the internal DB of InfluxDB):
I had the same issue, but the internal DB was not the fault, it was my own DB, as I was collecting much more data, my DB is 7,5GB big.
I have decided to spin up an LXC in Proxmox only for InfluxDB (tteck´s / community scripts) installed influxdb v1 and copied the data (the “influxdb” folder) via SFTP from HAOS under /mnt/data/supervisor/apps/data/a0d7b954_influxdb/
to the LXC under /var/lib/
Restarted InfluxDB, changed hass+glances configuration to point to new IP and done. All is working. Easier than to fiddle around with the container in hass..