Updated as branch v4.2.0-dev branch

Author: asraf-khan

aci-preupgrade-validation-script.py

@@ -38,7 +38,7 @@
 import os
 import re
 
-SCRIPT_VERSION = "v4.1.0-dev"
+SCRIPT_VERSION = "v4.2.0-dev"
 DEFAULT_TIMEOUT = 600  # sec
 # result constants
 DONE = 'DONE'
@@ -3010,17 +3010,19 @@ def scalability_faults_check(**kwargs):
 
 
 @check_wrapper(check_title="APIC Disk Space Usage (F1527, F1528, F1529 equipment-full)")
-def apic_disk_space_faults_check(cversion, **kwargs):
+def apic_disk_space_faults_check(cversion, tversion, **kwargs):
     result = FAIL_UF
     headers = ['Fault', 'Pod', 'Node', 'Mount Point', 'Current Usage %', 'Recommended Action']
     data = []
     unformatted_headers = ['Fault', 'Fault DN', 'Recommended Action']
     unformatted_data = []
     doc_url = "https://datacenter.github.io/ACI-Pre-Upgrade-Validation-Script/validations/#apic-disk-space-usage"
+    # we are checking /tmp utilization because high usage can lead to snaphshot corruption during an upgrade. After the fix version, snapshot storage location moved to /data.
     recommended_action = {
         '/firmware': 'Remove unneeded images',
         '/techsupport': 'Remove unneeded techsupports/cores',
-        '/data/log': 'Remove unneeded logs in var/log/dme/log'
+        '/data/log': 'Remove unneeded logs in var/log/dme/log',
+        '/tmp': 'Remove unneeded logs in /tmp directory'
     }
     default_action = 'Contact Cisco TAC.'
     if cversion.same_as('4.0(1h)') or cversion.older_than('3.2(6i)'):
@@ -3029,6 +3031,8 @@ def apic_disk_space_faults_check(cversion, **kwargs):
     dn_regex = node_regex + r'/.+p-\[(?P<mountpoint>.+)\]-f'
     desc_regex = r'is (?P<usage>\d{2,3}%) full'
 
+    tmp_faults_skip_versions = ["6.0(9f)", "6.1(4h)", "6.2(1g)"]
+    tmp_faults_skipped = False  # Track if we skip /tmp faults for CSCwo96334 versions
     faultInsts = icurl('class',
                        'faultInst.json?query-target-filter=or(eq(faultInst.code,"F1527"),eq(faultInst.code,"F1528"),eq(faultInst.code,"F1529"))')
     for faultInst in faultInsts:
@@ -3038,14 +3042,24 @@ def apic_disk_space_faults_check(cversion, **kwargs):
         fc = faultInst['faultInst']['attributes']['code']
         dn = re.search(dn_regex, faultInst['faultInst']['attributes']['dn'])
         desc = re.search(desc_regex, faultInst['faultInst']['attributes']['descr'])
-        if dn and desc:
-            data.append([fc, dn.group('pod'), dn.group('node'), dn.group('mountpoint'),
+        if dn:
+            mountpoint = dn.group('mountpoint')
+            # CSCwo96334: Skip /tmp faults when target is >= 6.1(4h) or any unaffected versions
+            if mountpoint == '/tmp' and (not tversion.older_than("6.1(4h)") or any(tversion.same_as(version) for version in tmp_faults_skip_versions)):
+                tmp_faults_skipped = True
+                continue
+            if desc:
+                data.append([fc, dn.group('pod'), dn.group('node'), dn.group('mountpoint'),
                         desc.group('usage'),
                         recommended_action.get(dn.group('mountpoint'), default_action)])
-        else:
-            unformatted_data.append([fc, faultInst['faultInst']['attributes']['dn'], default_action])
+            else:
+                unformatted_data.append([fc, faultInst['faultInst']['attributes']['dn'], default_action])
     if not data and not unformatted_data:
-        result = PASS
+        # If we only found /tmp faults that were skipped (CSCwo96334 fixed target versions), return NA
+        if tmp_faults_skipped:
+            result = NA
+        else:
+            result = PASS
     return Result(
         result=result,
         headers=headers,
@@ -3866,27 +3880,83 @@ def target_version_compatibility_check(cversion, tversion, **kwargs):
     return Result(result=result, headers=headers, data=data, recommended_action=recommended_action, doc_url=doc_url)
 
 
-@check_wrapper(check_title="Gen 1 switch compatibility")
-def gen1_switch_compatibility_check(tversion, fabric_nodes, **kwargs):
+@check_wrapper(check_title="Supported hardware compatibility")
+def supported_hardware_check(tversion, fabric_nodes, **kwargs):
     result = FAIL_UF
-    headers = ["Target Version", "Node ID", "Model", "Warning"]
+    headers = ["Target Version", "Node ID", "Model", "Type", "Warning"]
+    data = []
+    unformatted_headers = ["Target Version", "DN", "Model", "Type", "Warning"]
+    unformatted_data = []
     gen1_models = ["N9K-C9336PQ", "N9K-X9736PQ", "N9K-C9504-FM", "N9K-C9508-FM", "N9K-C9516-FM", "N9K-C9372PX-E",
                    "N9K-C9372TX-E", "N9K-C9332PQ", "N9K-C9372PX", "N9K-C9372TX", "N9K-C9396PX", "N9K-C9396TX",
                    "N9K-C93128TX"]
-    data = []
+    unsupported_6_0_1_switch_models = ["N9K-C93120TX"]
+    unsupported_6_1_1_switch_models = ["N9K-C93180LC-EX"]
+    unsupported_5_0_1_exp_module_models = ["N9K-M12PQ", "N9K-M6PQ", "N9K-M6PQ-E"]
+    unsupported_6_1_1_fex_models = ["N2K-C2332TQ-10GT", "N2K-C2348TQ-10GE", "N2K-C2232PP-10GE", "N2K-C2232TM-E-10GE", "N2K-C2348TQ-10G-E"]
+    unsupported_6_1_1_sup_models = ["N9K-SUP-A", "N9K-SUP-B"]
     recommended_action = 'Select supported target version or upgrade hardware'
-    doc_url = 'https://datacenter.github.io/ACI-Pre-Upgrade-Validation-Script/validations/#compatibility-switch-hardware-gen1'
+    doc_url = 'https://datacenter.github.io/ACI-Pre-Upgrade-Validation-Script/validations/#supported-hardware-compatibility'
 
-    if not tversion:
-        return Result(result=MANUAL, msg=TVER_MISSING)
-    if tversion.newer_than("5.0(1a)"):
+    if not tversion.older_than("5.0(1a)"):
         for node in fabric_nodes:
-            if node['fabricNode']['attributes']['model'] in gen1_models:
-                data.append([str(tversion), node['fabricNode']['attributes']['id'],
-                            node['fabricNode']['attributes']['model'], 'Not supported on 5.x+'])
-    if not data:
+            model = node['fabricNode']['attributes']['model']
+            if model in gen1_models:
+                data.append([str(tversion), node['fabricNode']['attributes']['id'], model, 'Switch', 'Not supported on 5.x+'])
+
+        eqptLCs = icurl('class', 'eqptLC.json')
+        for eqptLC in eqptLCs:
+            model = eqptLC['eqptLC']['attributes']['model']
+            if model in unsupported_5_0_1_exp_module_models:
+                dn = re.search(node_regex, eqptLC['eqptLC']['attributes']['dn'])
+                if dn:
+                    data.append([str(tversion), dn.group('node'), model, 'Expansion Module', 'Not supported on 5.x+'])
+                else:
+                    unformatted_data.append([str(tversion), eqptLC['eqptLC']['attributes']['dn'], model, 'Expansion Module', 'Not supported on 5.x+'])
+
+    if not tversion.older_than("6.0(1a)"):
+        for node in fabric_nodes:
+            model = node['fabricNode']['attributes']['model']
+            if model in unsupported_6_0_1_switch_models:
+                data.append([str(tversion), node['fabricNode']['attributes']['id'], model, 'Switch', 'Deprecated from 6.0(1)+'])
+
+    if not tversion.older_than("6.1(1f)"):
+        for node in fabric_nodes:
+            model = node['fabricNode']['attributes']['model']
+            if model in unsupported_6_1_1_switch_models:
+                data.append([str(tversion), node['fabricNode']['attributes']['id'], model, 'Switch', 'Deprecated from 6.1(1)+'])
+
+        eqptExtChs = icurl('class', 'eqptExtCh.json')
+        for eqptExtCh in eqptExtChs:
+            model = eqptExtCh['eqptExtCh']['attributes']['model']
+            if model in unsupported_6_1_1_fex_models:
+                dn = re.search(node_regex, eqptExtCh['eqptExtCh']['attributes']['dn'])
+                if dn:
+                    data.append([str(tversion), dn.group('node'), model, 'FEX', 'Deprecated from 6.1(1)+'])
+                else:
+                    unformatted_data.append([str(tversion), eqptExtCh['eqptExtCh']['attributes']['dn'], model, 'FEX', 'Deprecated from 6.1(1)+'])
+
+        eqptSupCs = icurl('class', 'eqptSupC.json')
+        for eqptSupC in eqptSupCs:
+            model = eqptSupC['eqptSupC']['attributes']['model']
+            if model in unsupported_6_1_1_sup_models:
+                dn = re.search(node_regex, eqptSupC['eqptSupC']['attributes']['dn'])
+                if dn:
+                    data.append([str(tversion), dn.group('node'), model, 'Supervisor', 'Deprecated from 6.1(1)+'])
+                else:
+                    unformatted_data.append([str(tversion), eqptSupC['eqptSupC']['attributes']['dn'], model, 'Supervisor', 'Deprecated from 6.1(1)+'])
+
+    if not data and not unformatted_data:
         result = PASS
-    return Result(result=result, headers=headers, data=data, recommended_action=recommended_action, doc_url=doc_url)
+    return Result(
+        result=result,
+        headers=headers,
+        data=data,
+        unformatted_headers=unformatted_headers,
+        unformatted_data=unformatted_data,
+        recommended_action=recommended_action,
+        doc_url=doc_url,
+    )
 
 
 @check_wrapper(check_title="Contract Port 22 Defect")
@@ -6292,7 +6362,7 @@ def multipod_modular_spine_bootscript_check(tversion, fabric_nodes, username, pa
 
     return Result(result=result, headers=headers, data=data, recommended_action=recommended_action, doc_url=doc_url)
 
-
+  
 @check_wrapper(check_title="Inband Management Policy Misconfiguration")
 def inband_management_policy_misconfig_check(cversion, tversion, **kwargs):
     result = PASS
@@ -6316,6 +6386,29 @@ def inband_management_policy_misconfig_check(cversion, tversion, **kwargs):
         result = FAIL_O
     return Result(result=result, headers=headers, data=data, recommended_action=recommended_action, doc_url=doc_url)
 
+    
+@check_wrapper(check_title="svccore excessive data check")
+def svccore_excessive_data_check(**kwargs):
+    result = PASS
+    headers = ['Class Name','Count']
+    data = []
+    recommended_action = "Delete the core files before proceeding with upgrade. Please refer to the document linked below and contact Cisco TAC for assistance if needed."
+    doc_url = "https://datacenter.github.io/ACI-Pre-Upgrade-Validation-Script/validations/#svccore-excessive-data-check"
+    try:
+        svccoreCtrlr_classes_count = icurl('class', 'svccoreCtrlr.json?query-target=self&rsp-subtree-include=count')
+        svccoreNode_classes_count = icurl('class', 'svccoreNode.json?query-target=self&rsp-subtree-include=count')
+        
+        if int(svccoreCtrlr_classes_count[0]['moCount']['attributes']['count']) > 240:
+           data.append(['svccoreCtrlr', svccoreCtrlr_classes_count[0]['moCount']['attributes']['count']])
+        if int(svccoreNode_classes_count[0]['moCount']['attributes']['count']) > 240:
+            data.append(['svccoreNode', svccoreNode_classes_count[0]['moCount']['attributes']['count']])
+        if data:
+            result = MANUAL
+        
+        return Result(result=result,headers=headers,data=data,recommended_action=recommended_action,doc_url=doc_url)
+    except Exception as e:
+        return Result(result=ERROR, msg="Error occurred while fetching svccore object counts: {}".format(str(e)), doc_url=doc_url)
+
 
 @check_wrapper(check_title='False Micron SSD failure_fault')
 def false_micron_ssd_failure_fault_check(cversion, tversion, **kwargs):
@@ -6428,7 +6521,7 @@ class CheckManager:
     api_checks = [
         # General Checks
         target_version_compatibility_check,
-        gen1_switch_compatibility_check,
+        supported_hardware_check,
         r_leaf_compatibility_check,
         cimc_compatibilty_check,
         apic_cluster_health_check,
@@ -6442,6 +6535,7 @@ class CheckManager:
         validate_32_64_bit_image_check,
         fabric_link_redundancy_check,
         apic_downgrade_compat_warning_check,
+        svccore_excessive_data_check,
 
         # Faults
         apic_disk_space_faults_check,
@@ -6693,4 +6787,4 @@ def main(_args=None):
         msg = "Abort due to unexpected error - {}".format(e)
         prints(msg)
         log.error(msg, exc_info=True)
-        sys.exit(1)
+        sys.exit(1)

docs/docs/validations.md

@@ -37,6 +37,8 @@ Items                                                        | This Script
 [Fabric Link Redundancy][g17]                                | :white_check_mark: | :no_entry_sign:
 [APIC Database Size][g18]                                    | :white_check_mark: | :no_entry_sign:
 [APIC downgrade compatibility when crossing 6.2 release][g19]| :white_check_mark: | :no_entry_sign:
+[Supported Hardware Compatibility][g20]                      | :white_check_mark: | :no_entry_sign:
+[Svccore Excessive Data Check][g21]                          | :white_check_mark: | :no_entry_sign:
 
 [g1]: #compatibility-target-aci-version
 [g2]: #compatibility-cimc-version
@@ -57,6 +59,8 @@ Items                                                        | This Script
 [g17]: #fabric-link-redundancy
 [g18]: #apic-database-size
 [g19]: #apic-downgrade-compatibility-when-crossing-62-release
+[g20]: #supported-hardware-compatibility
+[g21]: #svccore-excessive-data-check
 
 ### Fault Checks
 Items                                         | Faults         | This Script       | APIC built-in
@@ -267,6 +271,17 @@ The script checks the presence of generation one switches when the upgrade is cr
 Or you can check the [Release Note 15.0(1) of ACI switches][3] to see the list of generation one switches, typically the one without any suffix such as N9K-C9372PX, that are no longer supported from 15.0(1) release.
 
 
+### Supported Hardware Compatibility
+
+The script checks the presence of deprecated hardware in the fabric.
+
+The list of supported and unsupported hardware is populated from the Release Notes across all ACI releases. This means the check covers hardware compatibility changes introduced in any version, not just the most recent release. As new release notes are published and hardware is deprecated, this list is updated accordingly.
+
+Refer the [Release Note 15.0(1) of ACI switches][3] to see the list of unsuporrted hardware for your desired target versions. Prior upgrading to target version, replace the unsupported hardware elements in your fabric with other supported hardware.
+
+Contact cisco TAC for further assistance.
+
+
 ### Compatibility (Remote Leaf Switch)
 
 The script checks the requirement to use remote leaf switches on the target version.
@@ -2772,14 +2787,27 @@ Suggestion:
 Contact Cisco TAC to remove any identified misconfigured objects before performing the upgrade to prevent policyelem crashes.
 The [CSCwd40071][68] defect affects versions 5.2(5c) and later with a fix available in 6.0(1g). However, the issue will only be triggered during Apic upgrades crossing 6.0(4c) due to [CSCwh80837][67].
 
+
+### Svccore Excessive Data Check
+
+Due to excessive `svccoreCtrlr` or `svccoreNode` managed objects, Apic gui stuck in loading multiple queries.
+
+The svccoreCtrlr and svccoreNode objects represent core files related to Apic and Leaf/Spines process respectively.
+
+Due to [CSCws84232][67], the APIC GUI may become unresponsive after login, with dashboards stuck in a continuous “Loading…”state.
+Administrators may be unable to access or operate the APIC GUI, potentially impacting day-to-day management or upgrade.
+
+This check will verify the count of the `svccoreCtrlr` Managed Object and raise and alarm with the bug if object count found more than 240. Remove the content or objects of `svccoreCtrlr` or `svccoreNode`. Contact Cisco TAC or upgrade to a release containing the fix for CSCws84232 before proceeding with an upgrade.
+
 ### False Micron SSD failure_fault
 
-Due to [CSCwt38698][69], certain Micron SSDs present in the fabric may experience premature end-of-life failures after upgrading to `6.1(5e)` or `6.2(1g)`.
+Due to [CSCwt38698][70], certain Micron SSDs present in the fabric may experience premature end-of-life failures after upgrading to `6.1(5e)` or `6.2(1g)`.
 
 To avoid this issue, change the target version to another version. Or run the **SSD Lifetime Validation** script on all nodes with identified Micron SSDs prior to upgrading to determine the remaining SSD lifetime. If the SSD lifetime is critically low, you have to follow the SSD replacement procedure outlined in the field notice to ensure that the node remains available after the upgrade. In case you already upgraded your node and are experiencing unavailability due to this issue, contact Cisco TAC for the SSD replacement procedure to restore the node.
 
 - Script location: [SSD Lifetime Validation](https://github.com/datacenter/aci-tac-scripts/tree/main/SSD%20Lifetime%20Validation)
 
+
 [0]: https://github.com/datacenter/ACI-Pre-Upgrade-Validation-Script
 [1]: https://www.cisco.com/c/dam/en/us/td/docs/Website/datacenter/apicmatrix/index.html
 [2]: https://www.cisco.com/c/en/us/support/switches/nexus-9000-series-switches/products-release-notes-list.html
@@ -2849,4 +2877,5 @@ To avoid this issue, change the target version to another version. Or run the **
 [66]: https://bst.cloudapps.cisco.com/bugsearch/bug/CSCwr66848
 [67]: https://bst.cloudapps.cisco.com/bugsearch/bug/CSCwh80837
 [68]: https://bst.cloudapps.cisco.com/bugsearch/bug/CSCwd40071
-[69]: https://bst.cloudapps.cisco.com/bugsearch/bug/CSCwt38698
+[69]: https://bst.cloudapps.cisco.com/bugsearch/bug/CSCws84232
+[70]: https://bst.cloudapps.cisco.com/bugsearch/bug/CSCwt38698