x/stacktach
Code Issues Proposed changes

Remove select_related() from details/search/latest_raw calls

Browse Source 
The join caused by select_related() can cause mysql to make poor
choices in regards to index selection. It appears that in small to
mid sized deployments, searches on RawData when joined to
Deployment will use the index on the column being searched instead
of the 'when' that ultimately orders the data. This causes mysql
to create a temp table and file sort that table, which is quite
ineffecient.

This fix simply removes the select_related() calls and populates
the deployment field on the raws by hand.
This commit is contained in:
Andrew Melton 2013-10-30 12:50:35 -04:00
parent 221237366b
commit a126cfcb4b
1 changed files with 6 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
stacktach/views.py
View File

@ -367,12 +367,15 @@ def post_process_genericrawdata(raw, notification):
def _post_process_raw_data(rows, highlight=None):
deployments = models.Deployment.objects.all()
dep_dict = dict((dep.id, dep) for dep in deployments)
for row in rows:
if "error" in row.routing_key:
row.is_error = True
if highlight and row.id == int(highlight):
row.highlight = True
row.fwhen = dt.dt_from_decimal(row.when)
row.deployment = dep_dict[row.deployment_id]
def _default_context(request, deployment_id=0):
@ -412,7 +415,7 @@ def details(request, deployment_id, column, row_id):
c = _default_context(request, deployment_id)
row = models.RawData.objects.get(pk=row_id)
value = getattr(row, column)
rows = models.RawData.objects.select_related()
rows = models.RawData.objects
if deployment_id:
rows = rows.filter(deployment=deployment_id)
if column != 'when':
@ -448,7 +451,7 @@ def latest_raw(request, deployment_id):
c = _default_context(request, deployment_id)
then = datetime.datetime.utcnow() - datetime.timedelta(hours=1)
thend = dt.dt_to_decimal(then)
query = models.RawData.objects.select_related().filter(when__gt=thend)
query = models.RawData.objects.filter(when__gt=thend)
if deployment_id > 0:
query = query.filter(deployment=deployment_id)
rows = query.order_by('-when')[:20]
@ -469,7 +472,7 @@ def search(request, deployment_id):
updates = False
rows = None
if column != None and value != None:
rows = models.RawData.objects.select_related()
rows = models.RawData.objects
if deployment_id and int(deployment_id) != 0:
rows = rows.filter(deployment=deployment_id)
rows = rows.filter(**{column: value})