Polling an Oracle Linux 9.8 host on OCI (aarch64, 130.61.118.28, netOrk 0.46.0 on .50) times out after the 120 s poll budget, twice in a row (03:59 and 04:14 UTC on 2026-11-02). The poll worker log shows where the time goes:
04:14:50.693 Services: 203 service(s)
04:16:47.637 Poll timed out for device 3480a9a6-… after 120s
04:16:52.943 get_available_updates() failed:
Pattern not detected: '__NETORK_RC=\\d+\\s*\\n.*[#$\\>]\\s*$' in output.
get_available_updates() ran for its full 120 s read_timeout and never finished.
Why
_get_updates_rpm (napalm_linux/linux.py:1468) runs dnf check-update --quiet (yum check-update -q), and _rpm_security_names follows it with the security query (DNF_SECURITY_COMMAND from napalm-device-types, _YUM_SECURITY_COMMAND here), each with up to 120 s. Both download and parse fresh repository metadata whenever the cached copy has expired. On an OCI image (Oracle's repositories plus ol9_oci_included) on a small instance that alone outlasts the whole poll budget.
The apt path never does this: apt list --upgradable reads the local index. netOrk refreshes the index separately, once a day per host (refresh_available_updates → dnf makecache, NetOrk/netork jobs/tasks/updates/refresh.py), whose docstring states the contract: "A poll reads pending updates from the index the host has cached." The dnf and yum reads break it.
Expected
The poll reads only the cached metadata: dnf --cacheonly check-update, yum --cacheonly check-update, and the same for both security queries. A host with no cached metadata yet fails at once with a message that points at refreshing the index, instead of hanging until the timeout. The dnf security query is fixed in napalm-device-types (NAPALM/napalm-device-types issue filed alongside).
Priority: P2 — on an affected host every poll fails and the device is marked offline; any dnf/yum host whose metadata has expired pays the download inside its poll.
## What happens
Polling an Oracle Linux 9.8 host on OCI (aarch64, `130.61.118.28`, netOrk 0.46.0 on .50) times out after the 120 s poll budget, twice in a row (03:59 and 04:14 UTC on 2026-11-02). The poll worker log shows where the time goes:
```
04:14:50.693 Services: 203 service(s)
04:16:47.637 Poll timed out for device 3480a9a6-… after 120s
04:16:52.943 get_available_updates() failed:
Pattern not detected: '__NETORK_RC=\\d+\\s*\\n.*[#$\\>]\\s*$' in output.
```
`get_available_updates()` ran for its full 120 s `read_timeout` and never finished.
## Why
`_get_updates_rpm` (`napalm_linux/linux.py:1468`) runs `dnf check-update --quiet` (`yum check-update -q`), and `_rpm_security_names` follows it with the security query (`DNF_SECURITY_COMMAND` from napalm-device-types, `_YUM_SECURITY_COMMAND` here), each with up to 120 s. Both download and parse fresh repository metadata whenever the cached copy has expired. On an OCI image (Oracle's repositories plus `ol9_oci_included`) on a small instance that alone outlasts the whole poll budget.
The apt path never does this: `apt list --upgradable` reads the local index. netOrk refreshes the index separately, once a day per host (`refresh_available_updates` → `dnf makecache`, NetOrk/netork `jobs/tasks/updates/refresh.py`), whose docstring states the contract: *"A poll reads pending updates from the index the host has cached."* The dnf and yum reads break it.
## Expected
The poll reads only the cached metadata: `dnf --cacheonly check-update`, `yum --cacheonly check-update`, and the same for both security queries. A host with no cached metadata yet fails at once with a message that points at refreshing the index, instead of hanging until the timeout. The dnf security query is fixed in napalm-device-types (NAPALM/napalm-device-types issue filed alongside).
Priority: P2 — on an affected host every poll fails and the device is marked offline; any dnf/yum host whose metadata has expired pays the download inside its poll.
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
What happens
Polling an Oracle Linux 9.8 host on OCI (aarch64,
130.61.118.28, netOrk 0.46.0 on .50) times out after the 120 s poll budget, twice in a row (03:59 and 04:14 UTC on 2026-11-02). The poll worker log shows where the time goes:get_available_updates()ran for its full 120 sread_timeoutand never finished.Why
_get_updates_rpm(napalm_linux/linux.py:1468) runsdnf check-update --quiet(yum check-update -q), and_rpm_security_namesfollows it with the security query (DNF_SECURITY_COMMANDfrom napalm-device-types,_YUM_SECURITY_COMMANDhere), each with up to 120 s. Both download and parse fresh repository metadata whenever the cached copy has expired. On an OCI image (Oracle's repositories plusol9_oci_included) on a small instance that alone outlasts the whole poll budget.The apt path never does this:
apt list --upgradablereads the local index. netOrk refreshes the index separately, once a day per host (refresh_available_updates→dnf makecache, NetOrk/netorkjobs/tasks/updates/refresh.py), whose docstring states the contract: "A poll reads pending updates from the index the host has cached." The dnf and yum reads break it.Expected
The poll reads only the cached metadata:
dnf --cacheonly check-update,yum --cacheonly check-update, and the same for both security queries. A host with no cached metadata yet fails at once with a message that points at refreshing the index, instead of hanging until the timeout. The dnf security query is fixed in napalm-device-types (NAPALM/napalm-device-types issue filed alongside).Priority: P2 — on an affected host every poll fails and the device is marked offline; any dnf/yum host whose metadata has expired pays the download inside its poll.