Skip to main content

Crashes

3readings
1conditions
1themes fed
Livestatus

Windows and application crashes with decoded error details and local dump analysis when available.

Topic id: crashes.

How conditions work:one episode per stretch of trouble, rated by how bad it gets and how long it lasts.Device Health Conditions

Conditions (1)​

Select a condition for background, what to check and an example.

​

Critical
kernel_crash_recurring
What it is
Windows on this machine keeps crashing with blue screens.
Why it matters
Each crash restarts the machine and loses unsaved work, so users cannot rely on it, and a server keeps dropping every connection.
Display→Critical

Occurrences (2)​

​

kernel_crash
What it is
Windows stopped with a bug check (blue screen). 11 fields describe it
Why it matters
The machine restarted without warning: users lost unsaved work, and on a server every share, database and remote session dropped until Windows was back.

​

process_crash
What it is
A program on this machine crashed. 31 fields describe it
Why it matters
The program closed without warning: its user lost unsaved work, and a service it ran stopped until it was restarted.

Fields (144)​

One occurrence per crash, from the Windows event records that report it, with decoded codes and local dump analysis when a dump was available. The host row counts recent kernel crashes.

Full inventory every 8 hours. Supported changes are reported when the agent observes them.

Kernel crash counts (14)​

sparklogs.data.crashes.kernel_crash_count_1d
integer · countkernel_crash_recurring
Kernel crashes (stop errors) in the last day. Counts crashes, not dump files.Counts kernel_crash records; the latest is kernel_crash_last_ts.
sparklogs.data.crashes.kernel_crash_count_5d
integer · countkernel_crash_recurring
Kernel crashes in the last five days.Counts kernel_crash records; the latest is kernel_crash_last_ts.
sparklogs.data.crashes.kernel_crash_count_10d
integer · countkernel_crash_recurring
Kernel crashes in the last ten days.Counts kernel_crash records; the latest is kernel_crash_last_ts.
sparklogs.data.crashes.kernel_crash_count_30d
integer · countkernel_crash_recurring
Kernel crashes in the last thirty days, the whole history the agent keeps.Counts kernel_crash records; the latest is kernel_crash_last_ts.
sparklogs.data.crashes.kernel_crash_days_since_last
How long since the most recent kernel crash. Absent when there was none in the last thirty days.
sparklogs.data.crashes.kernel_crash_last_ts
string · timestamp
When the most recent kernel crash happened, the same time its kernel crash event carries. Until the agent has read the crash's memory dump, it can hold the estimated time, which moves to the time the dump records once the dump is read. Absent when there was none in the last thirty days.

Application crash counts (7)​

sparklogs.data.crashes.app_crash_count_1d
integer · count
Application crashes in the last day, all programs together. A crash near the start of the day can be left out.
sparklogs.data.crashes.app_crash_count_5d
integer · count
Application crashes in the last five days, all programs together. A crash on the first of the five days can be left out.
sparklogs.data.crashes.app_crash_count_10d
integer · count
Application crashes in the last ten days, all programs together. A crash on the first of the ten days can be left out.
sparklogs.data.crashes.app_crash_days_since_last
float · days
How long since the most recent application crash. Absent when there was none in the last thirty days.
sparklogs.data.crashes.app_crash_quiet_awake_days
float · days
How many days the machine has been awake without an application crash, counted from the newest one or from when the agent last started reading the Application log without a break, whichever is later. Sleep and low-power idle do not count. Absent when the agent cannot account for the machine's awake time.
sparklogs.data.crashes.apps_crashing_10d
integer · count
How many different programs crashed in the last ten days.

Crash record (13)​

sparklogs.data.crashes.crash_id
string
The crash's ID, the same on every report of the same crash.
sparklogs.data.crashes.time_basis
string
Where the event time comes from: dump_header (the time in the analyzed kernel dump), report_event (the crash's event log record), or estimated_bounds (no crash time is known; the time is time_upper_ts).
sparklogs.data.crashes.time_lower_ts
string · timestamp
For an estimated crash time: the earliest the crash could have happened, from the newest System log record before the following boot, or the latest crash evidence before that boot. A bound for investigating, not the crash time.
sparklogs.data.crashes.time_upper_ts
string · timestamp
For an estimated crash time: the earliest evidence known to come after the crash.
sparklogs.data.crashes.detected_ts
string · timestamp
When the agent first found the crash, which is not the crash time.
sparklogs.data.crashes.wer_report_ids
string_array
The Windows Error Reporting report IDs in the crash's records, lowercase: a report GUID, or on Windows Server 2012 R2 the small dump's file name, which the System report uses as its ID.

Failure and blame (5)​

sparklogs.data.crashes.blame_module
string
The first module on the crashing stack that is not part of Windows: where to look first in a debugger, not proof of the cause. For a kernel crash, the first driver that is not one of Windows' own, so a vendor driver is blamed. For a process crash, the first module outside the Windows folder (System32, SysWOW64 and WinSxS count as inside). Absent when no dump was analyzed or no frame qualifies.
sparklogs.data.crashes.blame_offset
string
Where in the blamed module the frame was, in hex: the faulting instruction for the first frame, a return address after it.
sparklogs.data.crashes.failure_name
string
What ended the crash. For a process crash, the .NET exception type when the runtime reported one, otherwise the name of the exception code, such as STATUS_ACCESS_VIOLATION. For a kernel crash, the stop error's name, such as IRQL_NOT_LESS_OR_EQUAL. Absent when no name is known.
sparklogs.data.crashes.failure_space
string
The code table the failure code belongs to: ntstatus for a process crash, bugcheck for a kernel crash. A .NET exception type in failure_name is not from that table. Present whenever the code is.
sparklogs.data.crashes.failure_code
string
The exception code of a process crash, or the stop code of a kernel crash, as 0x and 8 hex digits. Absent only when no record or dump carried one.

Dump analysis (13)​

sparklogs.data.crashes.analysis
string
How the dump analysis went: complete, partial_timeout, partial_error, resource_limited, dump_unavailable, unsupported_format or budget_skipped. complete means the analysis finished, not that the cause is known.
sparklogs.data.crashes.budget_reason
string
Why the dump was not analyzed. With budget_skipped: per_key (the limit for crashes of this kind in a day was reached), overall (the daily limit for all crashes was reached) or queue_full. With resource_limited: commit_pressure (the machine's committed memory stayed too high to start the analysis).
sparklogs.data.crashes.analysis_elapsed_ms
integer · milliseconds
How long the dump analysis ran. Absent when none ran.
sparklogs.data.crashes.analysis_peak_commit_bytes
integer · bytes
The most memory the dump analysis committed at one time. Absent when none ran or Windows did not report it.
sparklogs.data.crashes.matching_dump_count
integer · count
How many dump files matched the crash.
sparklogs.data.crashes.dump_path
string
The full path of the analyzed dump file. Absent when there was none.

Stack and modules (50)​

sparklogs.data.crashes.context_ip
string
The instruction pointer (rip or eip) at the crash, in hex.
sparklogs.data.crashes.context_sp
string
The stack pointer (rsp or esp) at the crash, in hex.
sparklogs.data.crashes.context_fp
string
The frame pointer (rbp or ebp) at the crash, in hex.
sparklogs.data.crashes.modules
object_array
The modules loaded at the crash, from the dump, at most 512; stack frames refer to them by position. Each has basename and, where read, machine (hex), size_of_image, time_date_stamp (8 hex digits), file_version, and the symbol identity a debugger needs: codeview_kind (rsds or nb10), pdb_name, pdb_guid (32 hex digits), pdb_signature (8 hex digits), pdb_age, portable_pdb, pdb_checksum_algorithm and pdb_checksum. The *_source members say where each part came from: module_stream, triage_entry, dump_image or local_image.
sparklogs.data.crashes.modules[].basename
string
In each modules entry.
sparklogs.data.crashes.modules[].machine
string
In each modules entry.

Kernel crash details (11)​

sparklogs.data.crashes.identity_basis
string
What identifies the kernel crash: boot_start (the boot it was reported in), report_id (its error report) or power_41 (the Kernel-Power 41 unexpected-restart record).
sparklogs.data.crashes.boot_start_missing
bool
True when no record shows when the boot that reported this kernel crash started.
sparklogs.data.crashes.bugcheck_code
string
The stop code, as 0x and 8 hex digits. Absent only when no record or dump carried one.
sparklogs.data.crashes.bugcheck_name
string
The stop error's name, such as IRQL_NOT_LESS_OR_EQUAL. Absent when the code has no known name.
sparklogs.data.crashes.bugcheck_space
string
The code table the stop code belongs to: always bugcheck. Present whenever the code is.
sparklogs.data.crashes.bugcheck_param1
string
The stop error's first parameter, in hex.

Process crash details (31)​

sparklogs.data.crashes.app_name
string
The crashed program's file name.
sparklogs.data.crashes.app_version
string
The crashed program's file version.
sparklogs.data.crashes.app_timestamp
string
The crashed program's build timestamp from its file header, 8 hex digits: a key for finding its symbols, not a date to read.
sparklogs.data.crashes.app_path
string
The full path of the crashed program.
sparklogs.data.crashes.pid
integer
The crashed process's ID.
sparklogs.data.crashes.process_start_ts
string · timestamp
When the crashed process started.

Example​

Inventory (every 8 hours)

crash history; kernel crashes in 10d 0, app crashes in 10d 0.

sparklogs.data.crashes.kernel_crash_count_1d: 0
sparklogs.data.crashes.kernel_crash_count_5d: 0
sparklogs.data.crashes.kernel_crash_count_10d: 0
sparklogs.data.crashes.kernel_crash_count_30d: 0

SparkLogs: CONTEXT, Info, crashes: INVENTORY: crash history; kernel crashes in 10d 0, app crashes in 10d 0.