Merge pull request #3201 from CumulusNetworks/create-pull-request/patch

Changes by create-pull-request action

Author: mergify[bot]

content/cumulus-linux-515/Whats-New/rn.md

@@ -14,6 +14,7 @@ pdfhidden: True
 
 |  Issue ID 	|   Description	|   Affects	|   Fixed |
 |---	        |---	        |---	    |---	                |
+| 4989428 | In rare cases, where a gNMI client terminates without cleaning up its gNMI gRPCs, the gNMI server might continue to buffer notifications for the client, resulting in an increase in gNMI server memory usage, potentially leading to a restart of the gNMI server. | 5.15.1-5.16.1 | |
 | 4986343 | PTM does not refresh certain entries and the PTM's neighbor status command (ptmctl -d) continues to show a neighbor that is already gone. This condition clears when the expected neighbor gets discovered. | 5.15.1-5.16.1 | |
 | 4986333 | LLDP session flaps might result in a PTMD process crash due to a double free memory block. | 5.11.2-5.16.1 | |
 | 4986329 | When many BFD sessions are configured at scale, <code>ptmd</code> might crash when one of the BFD sessions flaps. | 5.3.1-5.16.1 | |

content/cumulus-linux-515/rn.xml

@@ -7,6 +7,12 @@
 <th> Fixed </th>
 </tr>
 <tr>
+<td>4989428</td>
+<td>In rare cases, where a gNMI client terminates without cleaning up its gNMI gRPCs, the gNMI server might continue to buffer notifications for the client, resulting in an increase in gNMI server memory usage, potentially leading to a restart of the gNMI server.</td>
+<td>5.15.1-5.16.1</td>
+<td></td>
+</tr>
+<tr>
 <td>4986343</td>
 <td>PTM does not refresh certain entries and the PTM's neighbor status command (ptmctl -d) continues to show a neighbor that is already gone. This condition clears when the expected neighbor gets discovered.</td>
 <td>5.15.1-5.16.1</td>

content/cumulus-linux-516/Whats-New/rn.md

@@ -16,9 +16,13 @@ pdfhidden: True
 |---	        |---	        |---	    |---	                |
 | 4989440 | The <code>nv show interface qos-roce-status</code> command causes HIGH CPU. | 5.16.0-5.16.1 | |
 | 4989435 | During warm and fast reboots, NVUE intentionally skips a comprehensive Git cleanup to adhere to strict service start-up SLAs. When this cleanup is bypassed, existing Git maintenance debt, such as unreachable loose objects or blocked Garbage Collection (GC) states, remain in place and accumulate across successive reboots. The lightweight housekeeping performed during a standard <code>nv config apply</code> is not equivalent to a full cleanup. Consequently, stale objects and blocked GC states can persist until the repository becomes unreadable, eventually causing git commit operations to fail with the error <code>fatal: could not parse HEAD</code>.<br>To work around this issue, run <code>sudo systemctl restart nvued.service</code> manually after a warm or fast reboot to trigger the full cleanup path skipped during the automated startup sequence. | 5.16.1 | |
+| 4989434 | If the nginx <code>nvue.conf </code> file is updated (for example, with package update) and the NVUE git database is corrupt, the startup configuration might fail and the <code>nvue.conf</code> file might come from the debian package instead of the NVUE configuration that you create. To work around this issue, you can force the nginx <code>nvue.conf</code> file to update by making a basic change to system API, applying the change, then backing it out; for example:<br><pre>nv set system api compression gzip<br>nv config apply -y<br>nv unset system api compression gzip<br>nv config apply -y<pre> | 5.16.1 | |
+| 4989432 | When an interface is constantly flapping and the same BGP neighbor is coming up on a different interface, routes might not be reprogrammed in the kernel after the kernel deletes the routes during interface down events. To recover from this state, restart the FRR service. | 5.16.1 | |
 | 4989430 | If you add a static default route, then delete it, the switch removes the default route in the ASIC. | 5.16.0-5.16.1 | |
+| 4989428 | In rare cases, where a gNMI client terminates without cleaning up its gNMI gRPCs, the gNMI server might continue to buffer notifications for the client, resulting in an increase in gNMI server memory usage, potentially leading to a restart of the gNMI server. | 5.15.1-5.16.1 | |
 | 4989427 | When upgrading the switch to Cumulus Linux 5.16,  the upgrade process completes successfully and both gNMI and SNMP services are up, but gNMI telemetry data is not returned for several key metrics. | 5.16.0-5.16.1 | |
 | 4989426 | The <code>nv show interface qos-roce-status</code> command causes HIGH CPU. | 5.16.0-5.16.1 | |
+| 4989425 | Upgrading Cumulus Linux to 5.16.x causes a long boot delay and the <code>nvued</code> service remains in an activating state. To work around this issue, run <code>nvue-git-manual-recovery.sh</code> as a root user during maintenance on each node before upgrade, or on any node that has this issue. When successful, upgrade Cumulus Linux. If the script exits non-zero or reports time exceeded, follow the on-screen steps before continuing the upgrade. | 5.16.1 | |
 | 4986343 | PTM does not refresh certain entries and the PTM's neighbor status command (ptmctl -d) continues to show a neighbor that is already gone. This condition clears when the expected neighbor gets discovered. | 5.15.1-5.16.1 | |
 | 4986333 | LLDP session flaps might result in a PTMD process crash due to a double free memory block. | 5.11.2-5.16.1 | |
 | 4986329 | When many BFD sessions are configured at scale, <code>ptmd</code> might crash when one of the BFD sessions flaps. | 5.3.1-5.16.1 | |
@@ -246,6 +250,7 @@ pdfhidden: True
 |---	        |---	        |---	    |---	                |
 | 4989440 | The <code>nv show interface qos-roce-status</code> command causes HIGH CPU. | 5.16.0-5.16.1 | |
 | 4989430 | If you add a static default route, then delete it, the switch removes the default route in the ASIC. | 5.16.0-5.16.1 | |
+| 4989428 | In rare cases, where a gNMI client terminates without cleaning up its gNMI gRPCs, the gNMI server might continue to buffer notifications for the client, resulting in an increase in gNMI server memory usage, potentially leading to a restart of the gNMI server. | 5.15.1-5.16.1 | |
 | 4989427 | When upgrading the switch to Cumulus Linux 5.16,  the upgrade process completes successfully and both gNMI and SNMP services are up, but gNMI telemetry data is not returned for several key metrics. | 5.16.0-5.16.1 | |
 | 4989426 | The <code>nv show interface qos-roce-status</code> command causes HIGH CPU. | 5.16.0-5.16.1 | |
 | 4986343 | PTM does not refresh certain entries and the PTM's neighbor status command (ptmctl -d) continues to show a neighbor that is already gone. This condition clears when the expected neighbor gets discovered. | 5.15.1-5.16.1 | |

content/cumulus-linux-516/rn.xml

@@ -19,12 +19,34 @@
 <td></td>
 </tr>
 <tr>
+<td>4989434</td>
+<td>If the nginx {{nvue.conf }} file is updated (for example, with package update) and the NVUE git database is corrupt, the startup configuration might fail and the {{nvue.conf}} file might come from the debian package instead of the NVUE configuration that you create. To work around this issue, you can force the nginx {{nvue.conf}} file to update by making a basic change to system API, applying the change, then backing it out; for example:
+nv set system api compression gzip
+nv config apply -y
+nv unset system api compression gzip
+nv config apply -y</td>
+<td>5.16.1</td>
+<td></td>
+</tr>
+<tr>
+<td>4989432</td>
+<td>When an interface is constantly flapping and the same BGP neighbor is coming up on a different interface, routes might not be reprogrammed in the kernel after the kernel deletes the routes during interface down events. To recover from this state, restart the FRR service.</td>
+<td>5.16.1</td>
+<td></td>
+</tr>
+<tr>
 <td>4989430</td>
 <td>If you add a static default route, then delete it, the switch removes the default route in the ASIC.</td>
 <td>5.16.0-5.16.1</td>
 <td></td>
 </tr>
 <tr>
+<td>4989428</td>
+<td>In rare cases, where a gNMI client terminates without cleaning up its gNMI gRPCs, the gNMI server might continue to buffer notifications for the client, resulting in an increase in gNMI server memory usage, potentially leading to a restart of the gNMI server.</td>
+<td>5.15.1-5.16.1</td>
+<td></td>
+</tr>
+<tr>
 <td>4989427</td>
 <td>When upgrading the switch to Cumulus Linux 5.16,  the upgrade process completes successfully and both gNMI and SNMP services are up, but gNMI telemetry data is not returned for several key metrics.</td>
 <td>5.16.0-5.16.1</td>
@@ -37,6 +59,12 @@
 <td></td>
 </tr>
 <tr>
+<td>4989425</td>
+<td>Upgrading Cumulus Linux to 5.16.x causes a long boot delay and the {{nvued}} service remains in an activating state. To work around this issue, run {{nvue-git-manual-recovery.sh}} as a root user during maintenance on each node before upgrade, or on any node that has this issue. When successful, upgrade Cumulus Linux. If the script exits non-zero or reports time exceeded, follow the on-screen steps before continuing the upgrade.</td>
+<td>5.16.1</td>
+<td></td>
+</tr>
+<tr>
 <td>4986343</td>
 <td>PTM does not refresh certain entries and the PTM's neighbor status command (ptmctl -d) continues to show a neighbor that is already gone. This condition clears when the expected neighbor gets discovered.</td>
 <td>5.15.1-5.16.1</td>
@@ -1407,6 +1435,12 @@ hal_mlx_port.c:3919 ERR port state set failed for port &lt;#&gt;: Driver's Retur
 <td></td>
 </tr>
 <tr>
+<td>4989428</td>
+<td>In rare cases, where a gNMI client terminates without cleaning up its gNMI gRPCs, the gNMI server might continue to buffer notifications for the client, resulting in an increase in gNMI server memory usage, potentially leading to a restart of the gNMI server.</td>
+<td>5.15.1-5.16.1</td>
+<td></td>
+</tr>
+<tr>
 <td>4989427</td>
 <td>When upgrading the switch to Cumulus Linux 5.16,  the upgrade process completes successfully and both gNMI and SNMP services are up, but gNMI telemetry data is not returned for several key metrics.</td>
 <td>5.16.0-5.16.1</td>